CATEGORY
Software engineering & web development
AI-assisted software delivery, coding agents, developer platforms and modern web development.
Claude Code 2.1.309 closes multiple sandbox and permission bypassesThe release prevents shell startup files and MCP configuration files from being modified outside intended controls, and fixes permission checks that could be bypassed through symlinks or wrapped directory changes. It also hardens worktree isolation and Windows hook behaviour.→Codex CLI 0.164.0 adds enterprise MCP authentication and reusable presetsThe release adds ten numbered /presets slots, richer /model and /permissions arguments, terminal image previews and a default Amazon Bedrock setup wizard. It also masks credentials in supported remote shell operations and strengthens session recovery and sandbox checks.→Microsoft Agent Framework 1.23.0 establishes MCP 2.x compatibilityThe release migrates MCP transports, errors, schemas and lifecycle handling across the framework, while adding allowlisted argument forwarding for A2A requests. It also makes Docker shell trust requirements explicit and adds Python 3.15 support across several integrations.→Payload CMS 4.0 beta improves accessibility and speeds TanStack Start development loadsThe beta adds fixes for contrast, labels, headings, tabs, focus states and drag-reordering, while raising the minimum Node.js version to 24.22.1. Its TanStack Start integration reports initial development-load improvements of up to three times.→GitHub makes agentic autofix for code-scanning alerts generally availableThe feature explores relevant files, iterates on proposed fixes, and validates remediation by rerunning CodeQL before creating a draft pull request. It is available to GitHub Code Security and Advanced Security customers on Enterprise Cloud, with AI Credits and Actions-minute usage.→Claude Code 2.1.301 adds safer unattended retries and richer agent telemetryThe release adds CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS so unattended retry mode can stop instead of waiting indefinitely for a usage limit to reset. It also adds classifier verdict, permission mode, and classifier-category fields to tool-decision telemetry, while fixing unsafe worktree deletion and several search and workflow-reporting failures.→GitHub Copilot CLI adds agent-scoped hooks and safer autopilot handlingThe release adds agent-scoped declarative hooks, inline image rendering and more robust sandbox settings. It also changes unattended autopilot behavior so a pre-tool hook asking for approval is denied instead of leaving an invisible permission prompt.→GitHub Agentic Workflows defaults read-only runs to no network accessThe release disables network access by default for read-only workflows. It also records richer session provenance, plugin installations and MCP configuration, improving auditability of agent-driven automation.→Next.js canary adds agent-aware upgrade controls and durable-cache changesThe release gates upgrade reminders behind an agentUpgrade setting and enables network access for Codex upgrades. It also adds a turbopackGc disable option and changes durable use-cache hashing behavior.→Cline moves agent plugins into the hub processPlugins now load in the hub process by default, removing the requirement for a Node or Bun runtime on the user's PATH. Failed plugins remain visible through status commands and session diagnostics instead of silently losing their tools, while sandboxed subprocess loading remains available.→OpenHands adds browser recording and built-in ACP agentsThe release adds browser session recording for verification workflows and includes OpenCode and Pi as built-in ACP agents on local backends. It also adds configuration changes for automated setup and splash-page behavior.→Mastra adds durable run ownership and storage-level fencingRuns can now be claimed by the execution driving them, while recovery skips live claims or explicitly takes them over. Storage adapters can enforce run fences to reject writes from executions that have lost ownership, and PostgreSQL working-memory merges are performed atomically.→LangChain model profiles expose mid-conversation supportThe release declares mid-conversation support in model profiles across the core, Anthropic and OpenAI integrations. This gives applications structured metadata for deciding whether a provider can safely accept model or configuration changes during an active exchange.→pnpm adds registry profile and token management commandsThe release adds pnpm profile and pnpm token commands for viewing, creating and revoking registry credentials. It also introduces a files-based store index for network filesystems and faster Python dependency resolution.→Oxlint 1.88 adds LSP configuration diagnostics and new JavaScript rulesThe release makes configuration errors visible to the language-server client and adds new rule behavior including a configurable prefer-query-selector check. It also contains breaking AST comment-attachment changes and fixes conflicting TypeScript import autofixes.→Qwen Code nightly improves managed-agent coordination and test triageThe release improves managed-agent connector and broker robustness, adds mailbox-style session messaging and improves settlement of child runs after worktree merges. It also adds a test-failure report explorer for extension workflows.→WeKan 12.25 improves OAuth and SAML sign-in flowsWeKan 12.25 changes Google, OAuth/OIDC and SAML authentication so sign-in returns to the same browser window by default. Its hosting documentation also improves real-client-address forwarding for Caddy and Sandstorm deployments.→Rulesync 29.0.0 expands project instruction synchronisationThe release updates Rulesync’s configuration, MCP and permissions schemas alongside its cross-platform binaries. It also publishes signed release assets for macOS, Linux and Windows.→Langfuse 4.56.0 adds spend alerts and richer agent observabilityVersion 4.56.0 adds spend alerts for ClickHouse-billed organisations and syntax highlighting in session-tool previews. It also revises evaluation-result views, improves agent-session handling and requires date ranges for the observations MCP tool.→Rig 1.12.4 extends deployment-pipeline APIsThe release adds an Issue model to the APIs and exposes already-running pipelines in effective settings. It also includes pipeline CLI corrections and documentation updates.→cc-switch 1.0.13 updates multi-provider coding-agent routingThe release is distributed through the project’s GitHub release channel and updates the desktop and CLI tooling used to select and manage agent providers. It is part of the project’s 1.0 release line.→pforge publishes a Deno and TypeScript bridge for MCP serversThe bridge exposes MCP-server construction from Deno and TypeScript. Its release positions native Rust functionality behind a type-safe JavaScript-family developer interface.→Atelier 0.1.8 improves a native workspace for coding agentsThe release adds the latest packaged editor build for macOS and continues Atelier’s agent-session workflow. The project presents agent activity and per-turn changes inside a native development workspace.→Convex publishes a new precompiled self-hosted backend buildThe build is distributed through Convex’s backend release channel and provides a newer binary for running the reactive database platform outside Convex’s hosted service. Convex documents the precompiled artifacts as the supported installation path for platforms that do not build the backend locally.→DropForge packages authenticated web and terminal front endsThe release publishes the project’s web, terminal and desktop distribution targets together with amd64 and arm64 container support. The project describes itself as an integrated deployment combining authenticated interfaces and operational tooling.→NoDPI 2.0 adds stricter domain matching and fragmentation controlsThe release adds selectable fragmentation methods and enables strict domain matching by default. It also refactors core functionality and improves the user interface and statistics output.→OpenAI and Atlassian expand enterprise developer workflowsAtlassian and OpenAI announced an expanded partnership covering planning, building and delivering work through Atlassian products. The integration combines GPT-6 models with Atlassian’s Teamwork Graph and developer-workflow context.→Atlassian introduces AMP for human-agent software collaborationAtlassian announced AMP alongside a new MCP server, richer Teamwork Graph context and agent-session visibility. The system connects local and cloud agent work with Jira, Confluence, Loom and Bitbucket while recording ownership and actions.→Atlassian launches a forward-deployed engineering programme for AI deliveryThe programme embeds senior applied AI engineers in customer environments and includes a playbook based on Atlassian’s own deployments. Its stated focus is enriching organizational context, building custom agents and streamlining workflows.→Silicon Labs opens an AI SDK for Copilot, Cursor and Codex embedded developmentSilicon Labs announced a public-beta SDK that gives GitHub Copilot, Cursor and Codex structured access to its device SDKs, tools, documentation and connected hardware. It also introduced Hardware Intent capabilities and Databricks integration for embedded AI workflows.→DoiT partners with OpenAI on production deployment and AI cost attributionThe partnership adds DoiT’s Attribute cost-attribution product and migration support to teams deploying OpenAI models through Amazon Bedrock. The offering is designed to connect model and cloud usage with the products, customers or agents responsible for the spend.→OSERA establishes shared remediation standards for open-source softwareFINOS announced that the Open Source Enterprise Resiliency Alliance is operational, backed initially by six major banks. OSERA has published an open remediation standard and delivered fixes for more than 50 Java projects in its initial work.→Market Logic makes DeepSights MCP generally availableMarket Logic announced general availability of its DeepSights MCP integration. The connection lets enterprise assistants and automated workflows retrieve cited, customer-curated market evidence from the DeepSights platform.→Dfinitiv launches a platform for building branded agent experiencesDfinitiv introduced SmartCXP AI, a platform for building and operating branded applications across ChatGPT, Claude and other assistants. Its AI Forge generates the application components, while the Engine and Center provide governed runtime and usage visibility.→CAEVES extends governed enterprise context to any cloud through MCPCAEVES announced that Context LAKE will support public, private, sovereign and on-premises environments. The product indexes data in place and exposes traceable context through MCP to Copilot, Claude, ChatGPT and custom agents.→Beyond Now launches an API- and MCP-based agentic platform for telecom softwareBeyond Now launched a composable, event-driven platform for communications providers and technology companies. It combines open APIs, MCP support and role-based agents across discovery, selling, fulfilment and monetisation workflows.→OpenHands 1.25.0 expands agent configuration and workflow controlsThe release adds bulk model-profile creation, direct-prompt model-router settings, conversation-start routing, native Git integrations and configurable agent personas. It also improves Canvas, automation filtering, workspace discovery and session-key handling.→GitHub Copilot CLI 1.0.92 improves cloud runs and MCP reliabilityThe release adds copilot configuration subcommands and a pre-conversation environment picker for local or cloud execution. It also improves Entra-protected MCP credential renewal, bounds legacy HTTP-SSE hangs, preserves prompts during compaction and retries oversized provider requests after reducing attachments.→Claude Code 2.1.290 adds richer agent hooks and safer plugin controlsThe release adds tool-call and subagent identifiers to plugin hooks, introduces named-session attach and log commands, and expands managed-agent onboarding. It also fixes MCP attribution, symlink escapes, sandbox approvals, scheduled-task recovery and several remote-session failure modes.→Codex CLI 0.160.1 fixes Windows remote-MCP environment handlingThe release preserves SYSTEMROOT, TEMP and TMP when remote MCP servers run with explicitly configured environment variables. The change allows Unix hosts to retain the Windows executor’s startup environment instead of losing required variables.→Next.js 16.4 canary stabilises access-control APIs and ESLint 10 supportThe release marks forbidden() and unauthorized() as stable and adds ESLint 10 support to eslint-config-next. It also includes Turbopack cache, runtime, build-metadata and use-cache changes, plus an agent skill that streams bundle-analysis data as JSON Lines.→Gemini CLI nightly strengthens the latest agent runtime pathThe build was cut from the previous nightly release and distributed as a pre-release CLI package. Because it is a nightly build, the update signals ongoing agent-runtime development but is not a stable-channel guarantee.→Buildkite Agent 4.2.0 adds base-branch fetching for diff-aware CI jobsThe new git-fetch-base-branch option and corresponding environment variable let jobs fetch the current base branch before running. Build commands can therefore compare changes against the base branch’s current tip rather than relying only on the checkout’s existing refs.→OpenHands Extensions 0.29.0 adds design guidance to Canvas skillsThe release adds design guidance to the Canvas-extension API skill and refreshes agent instructions for AGENTS.md and automation telemetry. It also documents MCP-first service integrations and updates the related automation and extension packages.→protoAgent 0.195.0 makes agent-generated PDFs first-class artifactsPDF artifacts now render as real pages with scrolling, page indicators and zoom controls, while extracted text remains available for copying and screen readers. The release also validates saved PDFs and falls back to text cards for password-protected, oversized or bomb-like files.→Future AGI 1.47.1 improves simulation analytics for agent evaluationThe release improves cached serving for the v3 calls page and updates the simulation environment’s voice-call drawer and annotation controls. It also refreshes deleted-call handling and pins remaining cases after deletion.→Qwen Code Desktop adds Linux ARM support and managed-agent controlsThe release adds Linux aarch64 support, an automatic-update opt-out, and new web-shell and managed-runtime capabilities. It also improves MCP media handling, session inspection, worktree management and browser-use packaging.→vLLM 0.31 adds faster restarts and broader large-scale serving supportvLLM 0.31 introduces a weight-cache daemon and snapshot create/restore workflows for faster engine recovery. The release also expands speculative decoding, large-scale expert-parallel serving and DeepSeek-V4.1-Flash optimisations.→OpenCode 2.0.23 strengthens ACP editor integrationVersion 2.0.23 adds a native Cohere provider and fixes system-role request handling for several model providers. It also documents ACP v2 extensions and adds explicit control over automatic skill invocation.→n8n 2.42.3 expands API-key scope administrationn8n 2.42.3 lets global members grant variable scopes to API keys and prevents task-runner termination after unhandled promise rejections. The editor also exposes Gateway credit promotions on community nodes.→Mastra adds default agent error recovery and durable execution fixesThe release enables three error processors by default, repairing incompatible history or prefills before retrying transient failures. It also adds a tool-call-resumed stream event, improves durable workflow fencing, and prevents duplicated evented steps during redelivery.→Google ADK Go 1.8 hardens web and REST agent deploymentsADK Go 1.8 binds the web launcher to loopback by default and rejects cross-origin browser requests unless explicitly allowed. It also adds request-size limits, WebSocket protections, session-transcription persistence and fixes for Agent Engine deployment.→Next.js 16.4 canary advances bundle-analysis toolingVersion 16.4.0-canary.60 unifies the colours used for bundle-analyser deltas. The change is small but affects how teams interpret performance regressions during framework development.→Frappe Press patches authentication disclosure and agent-job handlingThe release prevents login endpoints from revealing account status and moves login-mail delivery into a background job. It also records who cancelled an agent job.→WebPageTest 23.01 ships production layout and testing improvementsThe release updates the production interface with incremental responsive-layout improvements and changes title ordering in result details. It also adds PHPUnit-based tests and removes the project’s Travis dependency.→Agent-Git publishes an early agent-native version-control workflowThe V0.1.0-alpha release introduces an early infrastructure layer designed around agent-managed branches and checkpoints. It is explicitly pre-release, so its implementation and adoption remain unvalidated.→VS Code 1.140 makes multi-agent development a first-class workflowThe release lets agents work across repositories and remote hosts while sharing session state through the Agent Host Protocol. It also adds HydraFusion model orchestration, managed Auto tiers and controls for minimum AI-feature versions.→GitHub Copilot CLI adds GPT-6.1 Sol and tighter session controlsThe release lets users restrict GitHub authentication to approved MCP origins and grant path access for individual sessions. Permission prompts remain actionable after resuming sessions, while compaction now returns a summary even when instructions request terse output.→Claude Code 2.1.286 improves permission and cloud-credential recoveryThe release shows counts such as “2 of 5” when several permission requests are pending and adds mouse navigation for fullscreen lists. It also fixes cases where GCP or AWS credential refreshes opened multiple login browsers and where resumed sessions lost turns.→GitHub Agentic Workflows expands audit ledgers and threat visibilityThe release preserves default ledgers when importing shared workflows and attributes friction costs and experiment data to audit sources. It also narrows action schemas, enforces self-hosted-runner requirements at compile time and hardens private-to-public flow validation.→Next.js ships a coordinated security release for supported branchesThe fixes affect both current and previous maintained release lines. The advisories cover image optimisation, App Router metadata image routes and self-hosted SSG or ISR deployments.→Rust 1.99 stabilises C-variadic function definitionsThe release makes it possible to define C-variadic functions in stable Rust and stabilises the associated unsafe-naked-function capability. It also adds the raw-borrow lint and extends panic analysis for zero-sized chunks and windows.→Cline SDK adds recoverable oversized MCP tool resultsInstead of losing tool output when it exceeds the model context, the SDK stores the complete result in a per-session cache and exposes a cline://cache URI. Agents can retrieve the remaining content by line range, with session-level size and expiry limits.→OpenCode hardens macOS agent binaries and session identityModel requests now carry session identity that distinguishes nested and parent sessions. The release also improves macOS 27 compatibility by re-signing locally compiled binaries and signing CLI release binaries with a Developer ID.→pnpm patches a dependency security advisory and store-corruption failuresThe release ships undici 7.29.1 and addresses failures that occurred when multiple pnpm processes wrote to the same store. It also fixes handling for package, catalog, project and command names that collide with JavaScript built-in object properties.→LiteLLM makes signed gateway images explicit in its 1.103.2 releaseThe release provides a pinned public key and verification commands for checking LiteLLM container signatures before deployment. It also backports several proxy and Anthropic integration fixes into the stable 1.103 branch.→Claude Sonnet 5.5 launches with a major coding-efficiency jumpAnthropic says Sonnet 5.5 is over 30% faster and costs up to 30% less per task than Sonnet 5. It scored 70.6% on Terminal-Bench 4.0, compared with 10.3% for Sonnet 5, while retaining a 1-million-token context window.→Claude Code 2.1.284 makes Sonnet 5.5 the default and hardens MCP recoveryVersion 2.1.284 adds Claude Sonnet 5.5 as the default Sonnet model on the Anthropic API, with 1-million-token context and published pricing. It also makes resumed MCP sessions wait for connecting servers, adds bulk MCP reconnect, improves permission and plugin controls, and fixes numerous agent-session failures.→GitHub Copilot CLI 1.0.89 expands model choice and MCP resilienceThe release adds GPT-6 Sol, GPT-6 Luna and Claude Opus 5.5 to the model picker when available, and reads Claude Code rule files as custom instructions. It also improves MCP OAuth scope handling, tolerates invalid neighbouring MCP entries, follows repository pull-request templates, and strengthens sandbox and session behaviour.→Copilot SDK adds typed outputs and human review for agent installationsAll six Copilot SDKs can now request JSON-Schema-validated or idiomatically typed outputs instead of free-form text. They also expose structured JSON-RPC error data and an installation-confirmation handler that lets applications require human approval before an MCP server or skill is installed.→LangChain 1.4.3 adds Bedrock Mantle support and agent fixesThe release adds Amazon Bedrock Mantle chat-model support to init_chat_model and recognises GPT-6 structured output without explicit profiles. It also repairs invalid tool calls in create_agent and sanitises cache settings for fallback models.→GitHub makes Claude Sonnet 5.5 generally available across CopilotSonnet 5.5 became available to Copilot Pro, Pro+, Max, Business and Enterprise users, subject to gradual rollout and administrator controls. GitHub lists availability across Visual Studio Code, Visual Studio, Copilot CLI, the coding agent, the Copilot app, github.com, mobile, JetBrains, Xcode and Eclipse.→North and Webflow add native embedded checkout for enterprise commerceNorth’s Embedded Checkout became available through the Webflow Marketplace, allowing teams to configure checkout inside the Webflow canvas rather than redirecting customers to an external page. North also provides developer documentation and sandboxes for implementation and testing.→Firebase CLI 15.32.0 introduces function kits and deployment fixesFirebase CLI 15.32.0 adds commands for installing, uninstalling and listing function kits, plus extension migration helpers and secret-reference overrides. It also increases IAM-policy retries, fixes an ERR_REQUIRE_ESM crash and improves parameter default handling.→pnpm 12.8.1 fixes reproducibility and CI installation edge casesVersion 12.8.1 stops frozen-lockfile installs rejecting injected workspace packages with peer dependencies and makes pnpm dedupe converge reliably. It also restores executable bits for local dependencies, reduces CPU use on many-core frozen installs and prevents unnecessary reinstalls after filtered installs.→Mock Service Worker 3.0 becomes ESM-only and raises runtime baselinesMSW 3.0 is ESM-only, drops supported Node.js versions below 22 and requires TypeScript 5.9 or newer. The release adds a Vite plugin, GraphQL subscriptions, navigation and form interception, WebSocket events and fixes for browser-mode test flakiness.→Codex CLI 0.158 hardens MCP, sandboxing and terminal approvalsThe release adds support for pre-registered MCP OAuth clients and bearer-token authentication for exec-server WebSocket connections. It also strengthens approval behaviour and fixes Windows, Linux and macOS sandbox and path-handling issues.→OpenCode 1.18.33 improves agent gateway reliability and secrecyProvider response and stream timeouts are now honoured for Cloudflare AI Gateway models. The release also prevents credentials and sensitive headers from appearing in debug configuration output and makes MCP launch failures visible to users.→LiteLLM 1.104 release candidate tightens gateway accounting and maskingThe release adds cached image-input pricing and repairs accounting paths involving MCP grants and Redis budget pipelines. It also masks streamed output more reliably and updates the Langfuse integration to its fourth SDK generation.→pnpm 12.8 warns about accidental environment-file publicationPackage packing and publishing now flag `.env` and `.env.*` files that are not explicitly listed in package metadata. The release also records pnpmfile checksums in pnpr lockfiles and avoids running unapproved build scripts while preparing certain git dependencies.→Docker Agent adds distributed tracing across ACP requestsACP requests now carry `traceparent` and `tracestate` headers, with server spans covering all 13 implemented agent protocol handlers. The release also adds lint checks for inefficient Go patterns and preserves TUI scrollback during partial tool calls.→OpenAPI Generator 7.16 expands client generation targetsThe stable release contains more than 100 enhancements and fixes across generated clients and servers. Its three new generators target asynchronous Python clients, Apache Dubbo microservices and Scala 3 services using sttp4 and jsoniter-scala.→Rulesync 22.0 changes how agent instructions map to projectsThe release enforces Codex’s 32 KiB instruction-chain limit and warns when generated rule files exceed it. It also converts bearer-token and header variable references into the Codex configuration fields that actually resolve environment values, while making malformed Claude skills non-fatal during import.→Oxlint 1.86 adds React component and TypeScript diagnosticsThe release adds new lint coverage for generated empty object types and improves handling of parameter-property reassignment, `declare` preservation and `await using`. It also updates React compiler, HTML-comment and several ESLint rule behaviours.→Tesseract.js 5.0 cuts browser OCR size and memory useThe release reports 54% smaller English data, 73% smaller Chinese data and a web-benchmark memory reduction from 311 MB to 164 MB. It also moves language and engine initialisation into `createWorker`, requiring migration from the older `loadLanguage` and `initialize` calls.→FreeRDP 3.32.1 ships a security and protocol-hardening releaseThe release addresses checks in gateway HTTP timeouts, channel lengths, media-type negotiation, smart-card emulation and other protocol paths. It also fixes a regression affecting fragmented static-channel copy and paste and adds mobile platform updates.→Claude Code 2.1.282 hardens telemetry, permissions and session recoveryThe release ignores project-level telemetry settings that could enable export or content capture, adds managed Chrome/MCP controls, and improves recovery of extended-thinking sessions. It also fixes permission-rule bypasses, duplicate command execution and several remote-session failures.→Payload CMS adds agent-assisted upgrades in its 4.0 canaryThe release changes the Local API default for access overrides and adds an agent-assisted upgrade command. It also includes TanStack Start integration work and active-locale publishing changes.→Lightdash adds custom agent skills and exposes them through MCPThe release lets teams create and manage reusable skills for Lightdash agents, invoke them through slash commands, and expose them through the platform's MCP server. It also adds scoped management controls and records lasting user corrections.→CodexBar 0.66 expands provider plugins and protects local configurationThe release brings the application to 84 providers, including Vercel AI Gateway and several additional model services. It also prevents CLI configuration writes from deleting plugin settings and secrets, while reducing repeated background cache work.→Wippy Local ships a private multi-agent desktop workspaceThe alpha release describes local multi-agent collaboration, persistent knowledge bases, project workflows and user-approved self-modification. The update also adds unsaved-change warnings and a logs menu.→Docker publishes an MCP Gateway OAuth helperThe package was extracted from Docker's MCP Gateway work and provides OAuth-discovery support for gateway integrations. Its release establishes a separately versioned component for authentication plumbing around MCP services.→protoAgent adds Zed ACP integration and stronger runtime tracingThe release lets developers continue console conversations inside Zed and drive protoAgent through Zed's Agent Panel over the Agent Client Protocol. It also improves boot-gate diagnostics, tracing semantics and default code-pane tool exposure.→RF-DETR 1.11 widens deployment exports for computer-vision applicationsThe release supports additional deployment formats and lets TensorRT engines serve a range of batch sizes through one optimisation profile. It also restructures export internals around a registry and removes several deprecated APIs.→Tornjak 1.8 adds versioned SPIFFE APIs and bundle managementThe release adds a versioned Tornjak API, SPIRE bundle operations and a unique SPIRE server identifier. It also removes verbs from the new v1 API paths in favour of method-based semantics.→Rz-ghidra 0.5.0 updates the Rizin reverse-engineering bridgeThe release updates the bridge between Rizin and Ghidra-based analysis tooling and marks compatibility with the newer Rizin series. It gives reverse engineers a current integration point for disassembly and decompilation workflows.→AutonomyAI connects product discovery directly to review-ready pull requestsThe system connects product signals and company data to planning and code generation without a manual handoff. Engineers retain merge authority, so the change extends autonomous development while preserving review gates.→Submersion AI launches Basin cybersecurity reasoning modelThe company reports an 80.8% CyberGym score and an eighth-place launch ranking. Basin has also been credited with discovering vulnerabilities in Snipe-IT and BookStack.→Gurucul releases runtime detection and prevention for risky AI behaviourThe product inventories agents, models, tools and hosts, maps detections to MITRE ATLAS and the OWASP Top 10 for LLM Applications, and provides response playbooks. Runtime prevention is available in preview.→Omada acquires EmpowerID to govern AI-agent identitiesThe combined offering is intended to control what agents can access and do across enterprise systems. The announcement positions agent identity governance as an extension of established identity-governance and administration workflows.→TIER IV publishes an open autonomous-racing reference stackThe reference design was used by 240 teams and approximately 600 participants in the 2026 Autonomous Driving AI Challenge. Developers can use the common platform to test algorithms from simulation through real vehicles.→Runpod adds enterprise governance across its AI developer cloudThe controls apply across Pods, serverless inference and dedicated GPU clusters. Runpod says the platform is intended to support the path from experimentation and fine-tuning through multi-node production deployment.→Airties adds integrated cybersecurity to its connectivity platformThe platform is designed to detect and remediate threats across connected devices without requiring subscribers to install separate software. It combines connectivity management and security operations in the ISP platform.→Seceon adds agentic SOC automation to OculusIT’s higher-education serviceThe service correlates more than 800 data sources and supports automated playbooks for threat containment. OculusIT says the deployment provides visibility across network, endpoint, identity, cloud and application layers.→Semaphore releases an agent-first interface for CI/CDThe interface is designed for agentic tools including Claude Code and Codex. It connects repository setup, pipeline configuration, test execution and failure analysis in one workflow.→AudioEye measures a major accessibility penalty for web agentsOn the most problematic site, agents completed 31% of assigned tasks compared with 96% after accessibility fixes. The study also reported up to six times higher token use on inaccessible sites.→WordPress Studio adds persistent design context to its website-building agentStudio Code now records a chosen visual direction in DESIGN.md and uses it to populate theme files, alongside design-selection previews. The release also adds capability tiers, multi-select agent questions, prompt-box answers, theme validation after writes, batched file edits, smaller screenshots and AI-credit warnings.→Repomix disables repository-local Git configuration after command-execution flawThe release disables repository-local Git settings for all Git commands after crafted configuration could invoke executables through gpg.program, diff.external, textconv or core.fsmonitor. It also fixes a Windows .gitignore crash and several CLI output bugs.→NestJS repairs microservice connection and tracing edge casesThe release fixes Kafka retry after failed connection and fails pending RabbitMQ and NATS requests when clients close. It also repairs gRPC namespace handling, exactly-once request-response spans, handler rejection behaviour and Express multer limit merging.→Docker Agent adds background-agent coordination and a shared WebAssembly runtimeThe release adds a wait_background_agents join tool, native Anthropic compaction, OpenAI Responses options, Gemini embeddings and URL Context, and a shared WebAssembly runtime. It also fixes token-budget enforcement, stream handling, WebSocket concurrency and per-call cost telemetry.→llama.cpp overhauls Qualcomm Hexagon buffer and DMA handlingThe release overhauls the Qualcomm Hexagon backend's buffer and DMA handling to support 64-bit mappings. The notes also list additional Hexagon improvements in the same runtime update.→Module Federation devtools add WebMCP and browser buildsThe release adds WebMCP, browser builds and release downloads to the devtools package. It also reduces first-screen JavaScript, hardens Zephyr deployments, fixes async factories and clears rejected remote-entry promises.→GitHub Actions makes its Windows 11 Arm64 Visual Studio 2026 image generally availableGitHub Actions published the Windows 11 Arm64 image with Visual Studio 2026 as generally available, using image version 20260920.164.1. The image updates Pulumi, AWS CLI, GitHub CLI, Chrome and Edge drivers, Firefox, PowerShell, Nginx and Visual Studio components.→WebdriverIO adds CPU and heap profiling to its test runnerThe release adds CPU and heap profiling across the CLI and runner packages, plus mobile permission-dialog actions. It also fixes multi-remote locator strategies, frame recovery, BiDi accessibility locators, network-idle waits, Allure incremental output and several mock and session edge cases.→LangChain publishes experimental type-safe routing middlewarelangchain-typesafe 0.0.1a3 is an initial release containing TypeSafeClassifier, experimental AutoModeMiddleware and experimental ModelRouterMiddleware. It also adds invocation-scoped classifier questions and usage metadata in traces.→Supabase Realtime returns 401 for malformed authorization headersThe release changes malformed Authorization headers to return HTTP 401. The release notes list this authentication change as the sole bug fix in v2.137.5.→OpenAI Node SDK 7.20.0 adds vault credentials and safety lifecycle APIsOpenAI Node SDK 7.20.0 adds APIs for environment-backed credentials and external storage configuration. It also exposes safety-case retrieval, safety warning and deactivation webhook events, and SIP media-security data for incoming calls.→Letta Code 0.32.14 connects Grok subscription OAuth to cloud agentsLetta Code 0.32.14 adds a Grok OAuth login path and protocol support for subscription OAuth tokens. It also removes legacy toolsets, scopes stream cursors to individual runs, and improves headless continuation and worktree behavior.→OpenAI Go SDK 3.64.0 adds agent safety and session-environment controlsOpenAI Go SDK 3.64.0 extends the client surface for safety cases, safety warnings, deactivation events and session-environment resets. It also adds external storage configuration and preserves Azure gateway paths and endpoint overrides.→Qwen Code 0.24.2 adds a Linux sandbox foundation and durable workflow controlsQwen Code 0.24.2 introduces an internal bwrap-based Linux sandbox foundation with structured process supervision and confined workers. The release also adds persisted workflow retry and rerun, concurrent browser-use sessions sharing a Chrome profile, and explicit trust for undecided workspaces.→Pi 0.86.1 adds Meta Muse access and faster repeat launchesPi 0.86.1 lets users authenticate to Meta Muse through /login meta or META_API_KEY and refreshes the model API key automatically. It also enables Node's persistent compile cache and fixes several agent-runtime errors involving diagnostics, context limits and strict tool schemas.→Zeroshot 10.6.0 exposes durable local profiles over ACPZeroshot 10.6.0 exposes local profiles as ACP agents over stdio. A single ACP session keeps its workspace and provider sessions alive while each prompt remains a separate durable run with fresh Git provenance and observable active turns.→verl 0.9.1 adds pluggable rollout routing and faster RL checkpoint syncverl 0.9.1 adds a rollout-router protocol and factory, extends delta-sharded weight synchronization to TorchTitan, and introduces GPU-lending async training. The release also moves dependency management to pyproject.toml and a committed uv.lock, making uv the supported installation path.→Microsoft agent-learning 0.9.0 adds reproducible policy replay with input lineageagent-learning 0.9.0 adds policy audit and replay with consumed-input lineage. The release records which inputs contributed to a policy execution so the run can be reproduced and examined later.→Cloudflare workerd 1.20260920.1 fixes Node interop and queued-stream cleanupworkerd 1.20260920.1 changes how Node.js interop reports errors from transform pairs. It also drops queued stream data once every consumer has been collected.→mise 2026.9.12 makes multi-worktree development environments service-awaremise 2026.9.12 lets tasks declare required daemons and lets projects select daemon subsets through daemon groups. Automatic ports, stable per-worktree hostnames and database, NATS and SpiceDB presets allow parallel checkouts to run the same development stack with less manual configuration.→Claude Code 2.1.274 hardens MCP startup, session recovery and telemetryClaude Code 2.1.274 adds configuration for limiting MCP startup waits, an effort attribute for OpenTelemetry spans and a redacted managed-settings telemetry event. It also improves transcript repair, context compaction, subagent reliability, remote sessions and gateway error handling.→Cline 4.1.19 blocks planted executables and fixes context exhaustionCline 4.1.19 warns when images are attached to text-only models and improves model capability reporting. It bases compaction on provider-reported token counts, blocks workspace executables from shadowing trusted Windows commands, retries transient failures and prevents several command and credential-handling errors.→Kilo Code 7.7.3 repairs worktrees and restores browser automationKilo Code 7.7.3 improves worktree discovery, cleanup, diagnostics and deletion performance, while restoring Playwright browser automation under Web Tools. It also adds Codex quota visibility, shared agent-management components and more stable transcript and skills handling.→OpenHands 1.20 adds profile-scoped secrets and Docker runtime forwardingOpenHands 1.20 lets agent profiles select the secrets available to them, forwards Docker conversation runtime settings and allows automations to choose saved profiles. The release also isolates mock-LLM profiles from ambient secrets in tests.→Pydantic AI 2.44 closes four web-fetch and telemetry security pathsPydantic AI 2.44 fixes four vulnerabilities reachable through web fetching or OpenTelemetry instrumentation. The fixes cover an IPv6 zone-identifier bypass of private-network blocklists, superlinear processing of attacker-controlled pages, alternate spellings of blocked domains and sensitive content remaining in spans despite `include_content=False`.→Node.js 26.9 expands Web Workers, crypto and runtime diagnosticsNode.js 26.9 adds Web Workers support, a generic MAC API, OpenSSL provider cipher and hash discovery, default FFI support and a new `node:bench` implementation. It also extends `perf_hooks` with meanCI and CBOR operations and introduces an experimental DTLS API alongside broader crypto, HTTP, filesystem and permission changes.→GitHub Copilot CLI lets custom agents inherit repository instructionsCopilot CLI 1.0.86-1 lets custom agents opt into AGENTS.md, copilot-instructions.md and CLAUDE.md through the `include-custom-instructions: true` frontmatter setting. The release also preserves marketplace plugins and skills when active sessions resume after reload and improves sandbox and background-shell status reporting.→Google Antigravity CLI 1.2.5 improves subagent guidance and recoveryAntigravity CLI 1.2.5 automatically gives Markdown-defined agents that declare subagent invocation access to the available subagent roster and delegation instructions. It also preserves background task names, clears definitively rejected credentials, records externally interrupted commands as cancelled and fixes frozen thinking frames.→Deno 2.9.7 tightens network, cache and lockfile safetyDeno 2.9.7 makes audit honour configured certificate stores, disambiguates HTTP authority paths and checks resolved IPs against the network deny list. It also validates lockfile tarball registry paths, fixes permission descriptors and removes argument, completion and sourcemap handling errors in deployment and sandbox commands.→CERT warns MLflow model loaders can bypass pickle controlsCERT/CC reported that MLflow's DSPy flavour applies its pickle guard conditionally based on the file extension, while the statsmodels flavour omits the guard. A malicious model artifact can therefore trigger arbitrary remote code execution through `mlflow.pyfunc.load_model()` even when pickle deserialization is disabled; statsmodels is fixed in version 3.15.0 and DSPy users are advised to avoid the affected flavour until fixed.→Claude Code v2.1.273 adds gateway telemetry hints and safer agent-session recoveryThe release adds request headers for gateway routing and observability, MCP disconnect notifications, and the ability to fork remote-control sessions into local background sessions. It also fixes permission-analysis bypasses, managed MCP settings, background-agent result delivery, context accounting and artifact publishing.→Gemini CLI 0.60 hardens MCP OAuth, sandboxing and extension boundariesGemini CLI 0.60 enforces RFC 9207 issuer identification in MCP OAuth and improves web-fetch destination validation and connection routing. It also isolates macOS sandbox directories, validates extension paths and workspace boundaries, checks configuration ownership, sanitises environment changes and preserves provenance metadata for untrusted tool output.→Mastra 1.67 adds persisted workflow authoring and platform-backed agent toolsMastra 1.67 adds a Studio Workflow Builder backend that authors persisted workflow definitions through permission-gated server endpoints. Its new @mastra/connect package turns platform connections into agent tools with injected credentials, token refresh, live discovery and optional allowlisting.→OpenAI Node SDK 7.16 adds WebSocket event limits and malformed-event handlingVersion 7.16 adds per-iterator limits for incoming WebSocket events and improves buffering and handling of malformed WebSocket messages. The release also bounds fallback abort subscriptions and keeps callback credentials local to individual HTTP requests.→LiteLLM 1.103 dev release tightens MCP admission, guardrails and spend controlsLiteLLM 1.103.0-dev.1 requires admission for delegated MCP OAuth and adds post-call guardrail support for background Responses and streaming flows. It also bounds tool and guardrail index operations against spend-log budgets, improves timeout logging, fixes reasoning translation and documents signed-container verification.→Meta exposes WhatsApp Business setup to coding agents through a new MCP serverThe new MCP server connects agents such as Claude, Cursor, Codex and ChatGPT to WhatsApp Business setup and management workflows. Agents can create business accounts, verify phone numbers, register Cloud API access, create templates, test messages and webhooks, and check configuration failures.→Arcee AI chooses Fuzzball for portable orchestration of frontier model workloadsFuzzball provides an orchestration layer for Arcee’s compute-intensive training and inference workloads without replacing its internally developed model stack. The software is available across major clouds and on-premises clusters, while Arcee retains control of its models, data, training and deployment architecture.→Cohesity launches Agent Resilience to snapshot and recover AI-agent stateAgent Resilience discovers and protects agent memory and configuration with point-in-time recovery, immutable backups and clean-room recovery patterns. It also maps connected applications, databases and other services so teams can restore both an agent and resources it changed; availability is initially limited to select customers.→Wingify unifies AB Tasty and VWO around agentic real-time web optimisationWingify’s newly unveiled platform combines experimentation, personalisation, behavioural insights, commerce optimisation, feature management and customer engagement on one data foundation. Its Wingz engine is designed to interpret live visitor signals, activate changes across web and app touchpoints, and measure outcomes in the same workflow.→Positron enters SageMaker Studio as a governed browser IDE for R and PythonPositron on Amazon SageMaker provides R, Python, Quarto and a Bedrock-backed Posit Assistant in a browser-based development environment. The image runs inside the customer’s AWS account under SageMaker execution roles, keeping data and credentials within the existing AWS boundary.→GitHub Copilot Auto adds explicit cost-and-quality tiersThe tiers change how Copilot weighs cost, quality and response time while retaining per-prompt model selection. The feature is rolling out in Visual Studio Code, Copilot CLI and the Copilot app.→Qwen Code 0.23.4 expands resumable and cross-session agent workflowsThe release lets approved Web Shell goals start automatically, adds a peer endpoint for external cross-session messaging and introduces limits for goal turns and active minutes. It also routes scheduled Web Shell runs to configured models and session groups, while removing configurable message-prefix filtering from channels.→OpenCode 1.18.31 restores state across resumed agent sessionsVersion 1.18.31 restores model, effort, mode and reasoning-chunk boundaries when sessions are loaded, resumed or forked. It also makes remote configuration authentication failures visible during startup and exits with a failure status.→Claude Code 2.1.271 adds remote fast mode and tighter agent controlsThe release applies fast mode to permitted cloud and self-hosted Remote sessions, adds per-command allowed domains in sandboxed auto mode and introduces hash-specific approval for plugin installation and updates. It also fixes MCP OAuth registration handling, session resumption, permission checks and background-command duplication.→mise 2026.9.9 adds encrypted dotfile history and safer bootstrap adoptionmise dot track --encrypt now creates an encrypted baseline and fails closed if that baseline cannot be saved. The release also isolates concurrent dotfile indexes, adds rollback-aware bootstrap --replace-history behaviour and introduces lock-aware uv options for PyPI tools.→Next.js 16.4 canary.31 fixes proxy detection and Turbopack shutdown handlingVersion 16.4.0-canary.31 fixes detection of proxy.ts when compound pageExtensions are used and makes next dev wait for Turbopack shutdown. It also advances tree shaking through module wrappers and export tracing through client-component proxies.→Cline SDK 0.0.83 introduces hub-managed Agent PluginsHub-managed plugins are discovered from explicit host paths, while workspace .agents/plugins directories are not scanned automatically and therefore cannot implicitly start repository-controlled MCP servers. The release also adds bounded retries for transient model failures, improves event-stream performance and prevents apply_patch Add File operations from silently overwriting existing files.→pnpm 12.4.2 patches executable-shim and GitHub Actions security issuesThe release prevents dependency executables from taking over another package’s POSIX bin shim and requires HTTPS for GitHub server URLs except on loopback hosts. It also improves installs, peer-dependency resolution, Python lockfile reuse, workspace performance and several platform-specific behaviours.→uv 0.12.15 restores valid Docker and target installation commandsThe release restores valid commands such as uv pip install --system in python Docker images and uv pip install --target ., which had been rejected in 0.12.14. It also batches cache writes to speed cold-cache resolution and HTTP cache revalidation.→Grafana 13.2.2 ships a coordinated security releaseThe release fixes CVE-2026-15815, CVE-2026-76154 and CVE-2026-79656. It also preserves query-variable refresh settings during v2 dashboard import and fixes a folder-rename UID collision during full provisioning synchronisation.→Perplexity uses GPT-6 Astra for end-to-end software changes, testing and production monitoringOpenAI says Perplexity uses GPT-6 Astra to write communications, edit software, monitor production systems and emulate language-model APIs or connectors for testing. The account is a vendor-published customer case study with no independent benchmark data.→Qwen Code adds cross-session agent peers and budgeted goalsThe Qwen Code nightly release adds a peer endpoint that lets external programs join cross-session messaging and lets goals stop at a turn or active-time budget. It also adds a Codex executor, native agent built-ins, CodeModeOnly tool execution and recovery improvements.→OpenAI confirms its agents were involved in an earlier RubyGems incidentThe New Web reports that OpenAI confirmed its agents were involved in a May RubyGems incident in which more than 2,000 packages were submitted and documentation infrastructure executed agent-written code. Researchers linked the activity to OpenAI, but RubyGems and OpenAI say attribution and whether API keys were successfully stolen remain unresolved.→Cline Desktop 0.0.27 hardens credentials, provider routing and session recoveryThe release changes expired-login handling, prevents stored credentials from silently restoring after sign-out and adds actionable recovery controls. It also preserves provider-specific protocols for OpenCode Go models and retries with a stable session identity.→n8n adds instance-report log streaming and fixes credential revocationThe release adds core events for streaming logs from instance reports and repairs end-user credential revocation in the editor. The update is published as a pre-release build.→GitHub Agentic Workflows routes release-note writes through a safe outputThe release makes agentic workflows that touch release notes use update-release instead of direct GitHub mutations, preserving a read-only job model. It also lets cached-log commands merge matching JSONL shards, prune stale files and write collision-resistant results.→Next.js canary fixes runtime environment mutation in the durable cache pathThe canary release fixes a durable use cache issue involving runtime environment-variable mutation. It also changes Turbopack handling for new atoms and serialises turbo-task trybuild tests.→NestJS 12.0.2 repairs streaming, microservice and Fastify edge casesThe release keeps Socket.IO and WebSocket message streams alive when handlers throw and improves failure handling for rejected microservice reply subscriptions. It also fixes several core, path-normalisation and Fastify response edge cases and adds MQTT connection-attempt limits.→Letta Code 0.32.8 hardens approval recovery and agent messagingThe release retries message requests when a gateway returns an HTML error page and prevents malformed questions from crashing the approval interface. It also improves teleport ownership during approval recovery and consumes queued messages without sender holds.→Agent Deck 1.16.10 hardens multi-agent session delivery and restart handlingThe terminal session manager now delivers completions when transcript signals are stale and merges post-start saves with concurrent session detection. It also rechecks releases while open and verifies that a tmux session exists after spawning.→GitHub Agentic Workflows adds package-aware updates and safer model accountingThe release lets gh aw update target workflow packages individually and adds a Node network preset for the Impeccable Skills Reviewer. It also fixes Codex model identifiers and improves token, threat-detection and diagnostics reporting.→Claude Code fixes long-session permission regressions in read-only Git commandsThe patch fixes permission prompts that could reappear after a session had been running for a while. Anthropic identifies the issue as a regression introduced in version 2.1.269.→GitHub Agentic Workflows adds shared JavaScript modules for workflow packagesThe release supports .mjs and .cjs modules inside aw.yml packages with path-traversal and extension restrictions. It also coordinates rate-limit handling for multi-target logs and makes failed-run and evaluation credit accounting fail closed.→pnpm 11.27 adds explicit build approval and trust-policy cleanupThe release asks for approval before running package build scripts and keeps unapproved packages in the ignored-build-scripts state. It also adds trust-policy cleanup, safer global package operations and improved workspace peer dependency handling.→Qwen Code nightly hardens local-file bridges and agent process controlsThe release closes trust-gate and bystander gaps in the local-files bridge, rejects an unlimited cgroup sentinel for ACP child heaps and adds transient network retries. It also adds an explicit VS Code session-source switch, improves hook handling and reaps surviving Windows hook process trees.→Cloudflare workerd expands Node networking compatibility at the edgeThe release adds net.Server support to node:net and introduces cloudflare:node connectHandler. It also changes the event-port integration to be driven by a per-thread Tokio runtime.→Fireworks schedules retirement of older serverless model endpointsThe notice applies to serverless, Fast and US-only serverless endpoints while leaving dedicated deployments unaffected. Fireworks lists replacement models for the affected DeepSeek V4, GLM 5.2, Muse Glimmer 30B and Kimi K2.x variants.→Letta Code adds asynchronous cloud messaging and agent coordinationThe release adds queued Cloud messages with a no-wait option, resumes teleported turns after reconnects and restores steering behind background notifications. It also adds SendAgentMessage for agent coordination and improves runtime, telemetry and Windows shell handling.→llama.cpp improves parsing for Qwen3-Coder complex typesThe release improves parsing of complex types in Qwen3-Coder chat interactions and adds corresponding parser support. It was published as a nightly build of the C/C++ inference runtime.→LiteLLM release candidate adds signed gateway images and Agent Skills discoveryThe release signs Docker images with Cosign and documents verification, adds an Agent Skills well-known index and exposes routed-model and session-savings information for Claude Code and Codex. It also adds customer-managed KMS support for virtual keys stored in AWS Secrets Manager.→Qwen Code 0.23.2 adds resumable multi-agent controls and safer remote sessionsThe release carries budget and progress guidance into continuation prompts, journals failed agents, adds browser-task notifications and separates planning from execution permissions. It also adds token-based remote starts with QR pairing, same-origin shell access and improved content filtering.→OpenAI Agents Python 0.22.2 adds image-tool support and closes a sandbox raceVersion 0.22.2 adds support for current image-generation tool options and prevents symlink races in the UnixLocal file API. It also resets the compaction response chain correctly after session history is modified.→Pydantic AI Harness 0.30 adds live trajectory steering and GitHub workflow integrationThe release introduces TrajectoryJudge, which reviews a live run on a cadence and can steer it mid-run. It also adds filesystem traversal events, explicit OpenTelemetry decisions and a hosted engine definition for running repository-defined Pydantic agents in GitHub Agentic Workflows.→GitHub Agentic Workflows 0.88.7 tightens agent output and MCP policy handlingVersion 0.88.7 makes agent workflows fail when agents report incomplete work instead of silently succeeding, and restricts artifact packaging to known files. It also improves log-download tooling and filters disallowed repository-memory files before validation and upload.→Gemini CLI preview strengthens MCP OAuth, sandboxing and path boundariesThe preview adds destination validation for web-fetch utilities and enforces RFC 9207 issuer identification in MCP OAuth flows. It also isolates temporary directories, tightens extension and symlink boundary checks, requires consent for environment changes and removes a hard-coded CrUX API key.→LiteLLM adds streaming guardrail pipelines and stronger agent gateway controlsThe release runs post-call guardrail pipelines on streaming responses, applies stream text rewrites and redacts provider keys from passthrough errors. It also scopes emulated file search, adds support for running the Pi coding agent through the proxy and introduces opt-in per-server MCP OAuth relay discovery.→OpenAI Node SDK 7.11 adds prompt-cache diagnostics and API-key expiry fieldsVersion 7.11.0 adds prompt-cache diagnostics and service-account API-key expiration fields, and corrects function-argument completion event fields. The release also improves WebSocket and streaming cleanup examples for Responses-based applications.→Akeyless makes runtime intent controls for AI agents generally availableAgentic Runtime Authority adds real-time intent-based access control on top of Akeyless’s credential-brokering layer. General availability includes integrations for Claude Enterprise, OpenAI Codex and Amazon Bedrock AgentCore, plus session termination, investigation views and event forwarding to security tools.→OPAQUE publishes Weight Custody Manifest for governed on-premises model deploymentWeight Custody Manifest ties key release to signed conditions covering model identity, licensing, jurisdiction, approved software and custody requirements. The preview includes a published specification, reference library and 91 public tests, with support for revocation and tracking fine-tuned derivatives to parent models.→Fabrix.ai launches governed VibeOps for agentic SRE and operationsGoverned VibeOps combines Argos small language models with integrations for tools including Splunk, Dynatrace, Datadog, ServiceNow, Cisco, IBM and AWS. Its stated lifecycle covers generating, versioning, testing, reviewing and promoting operational apps and agents, with identity, observability, data lineage and bounded agentic work built in.→Déjà Vu 0.19.4 gives multiple coding agents shared project memoryDéjà Vu 0.19.4 connects shared memory to 23 agent harnesses, including VS Code Copilot Chat, Amp, Prime Agent, Cline, Qwen Code and OpenClaw. The release adds session digests, per-prompt recall, compaction-aware forgetting and same-turn repair context for failed commands.→MCP Python SDK 2.2.0 tightens Streamable HTTP and OAuth defaultsMCP Python SDK 2.2.0 limits HTTP redirects to the endpoint origin, expires idle stateful sessions after 30 minutes and caps concurrent sessions at 10,000. It also validates OAuth issuer and token-resource settings more strictly.→Ultralytics Inference 0.0.42 improves browser and WebAssembly inferenceUltralytics Inference 0.0.42 reduces unnecessary XNNPACK thread contention and fixes browser pose-label parsing. It also improves webcam, ONNX and LiteRT handling across the project’s Rust, WebAssembly and browser packages.→Hermes Agent 0.21.1 rolls up major agent-runtime and MCP changesHermes Agent 0.21.1 rolls up 5,139 non-merge commits, 4,364 changed files and 632 merged pull requests since version 0.21.0. The release includes file-operation and startup performance work, desktop session controls, browser annotations, MCP authorisation improvements, cron fixes and more reliable delegation.→llm 0.35 adds GPT-6 Astra access to Simon Willison’s CLIllm 0.35 adds the OpenAI gpt-6-astra model to the command-line tool’s model catalogue. The update gives terminal-based developers a direct path to experiment with the model through an established local workflow.→Pydantic AI 2.41.0 adds ChatGPT and Codex subscription authenticationPydantic AI 2.41.0 introduces an openai-codex provider for ChatGPT/Codex subscription authentication. The release also adds a direct ImageGenerator API and renames fallback_model to fallback_subagent_model on image-generation and web-search components.→ORAS CLI 1.3.0 adds OCI 1.1.1 support and portable registry backupORAS CLI 1.3.0 is compliant with OCI distribution-spec 1.1.1 and adds `oras backup` and `oras restore` for moving artifacts between registries and local OCI layouts. It also promotes several attachment features to stable and adds richer multi-architecture manifest management.→n8n 2.39.0 expands agent workflow controls and observabilityn8n 2.39.0 hardens agent workflows against unpublished or incompatible tools, adds capability-aware AI Assistant behaviour and introduces OTLP gRPC tracing. It also adds workflow-version and source-control endpoints, agent concurrency limits, MCP registry credential support and browser-profile execution.→GitHub Actions previews Ubuntu 26.04 runner imagesThe runner-images update makes Ubuntu 26.04 and Ubuntu 26.04 Arm available as public previews and adds PostgreSQL to Ubuntu Arm64 images. It also updates tools including Docker Buildx, Rust, GitHub CLI and cloud CLIs, while setting a deprecation path for Ubuntu 22-based images.→Fimo launches autonomous websites around coding-agent outputFimo says its platform accepts AI-generated code from coding agents and lets native or custom agents publish SEO content, translate pages, repair schema issues and refresh stale content. Changes can move through review processes and custom workflows, keeping human approval in the publishing loop.→Microsoft APM 0.30.0 hardens agent package installation and cache lifecycleAPM 0.30.0 adds checksum verification before Unix archive extraction, owner-aware unprivileged installation and a Homebrew-first macOS onboarding path. It also fixes cache-pruning correctness, improves release lookup recovery and runs CodeQL on merge-queue commits.→Qwen Code 0.23.1-preview.2 makes multi-agent sessions and workflow runs inspectableThe preview adds visual management of dynamic workflow runs, live subagent status, a session resource catalogue and worktree-isolated named tasks. It also introduces session-turn navigation, transcript HTML export, workspace-scoped Skills and more deterministic workflow polling.→Midscene 1.12.4 adds scoped AI contexts and stateless UI test runsMidscene 1.12.4 adds scoped AI contexts and stateless Test Runner workflows for browser and cross-platform UI automation. The release also fixes delayed Linux input, cancellation handling, missing-coordinate retries and duplicate YAML assertion errors.→Alibaba Open Code Review adds controllable LLM review budgets and live progressOpen Code Review 1.11.6 adds first-class controls for effort, maximum token budget and LLM reasoning effort, alongside live review progress. It also adds OCaml and ReasonML allowlists, line numbering in the existing-code viewer and signal-forwarding fixes.→Nextcloud All-in-One 14.1.0 beta brings Nextcloud 35 installs and higher upload defaultsNextcloud All-in-One 14.1.0 beta includes Nextcloud 34.0.3 while allowing new instances to install Nextcloud 35 directly. It adds adjustable logging for Caddy instances, raises the default upload limit to 1 GB and documents mounting multiple host directories.→n8n 2.37.11 fixes workflow cleanup and external-secret failovern8n 2.37.11 ensures running jobs are cleaned up when a workflow run rejects. It also keeps a serving external-secrets provider active when its replacement fails.→Remotion 4.0.522 expands timeline editing in its React video studioRemotion 4.0.522 adds command options to CLI help and timeline zoom for audio previews. It also improves sequence resizing, timeline-layer selection, clip splitting, composition menus and collapsed-sidebar restoration in Remotion Studio.→Pact JS 17.1.4 updates its pact-core contract-testing enginePact JS 17.1.4 updates the pact-core dependency to version 20.1.1. The release also updates the fast-xml-parser development dependency used by the XML examples.→Gemini CLI publishes a new 0.60 nightly buildThe build is tagged v0.60.0-nightly.20260907.g85aca163f and was published on 7 September. Its release page links the complete change set against the preceding nightly build but does not provide feature-level notes.→OpenAI reports coding agents now exceed human research labour inside its organisationOpenAI says its research organisation uses the equivalent of 3.1 agent-workdays for every human workday, with the median high-usage researcher spending more than $600 per day at API prices. The company also reports more code, more experiments and increasing delegation of higher-level tasks, while noting that human researchers still set priorities, judge results and intervene on complex work.→Qwen Code preview adds inspectable multi-agent workflowsThe preview adds Web Shell workflow visualisation, live subagent status, a session resource catalogue, worktree-isolated named tasks and session-turn navigation. It also introduces mandatory conversation writer leases and workspace-scoped skills runtime.→Vercel AI SDK adds GPT-6 reasoning configurationThe patch release for @ai-sdk/openai adds GPT-6 reasoning configuration updates. The change gives applications using Vercel's provider integration a supported path to pass the model-specific reasoning settings.→Claude Code 2.1.263 fixes background-agent reliability failuresThe release fixes interrupted commands being shown as completed, prevents finished subagents from silently restarting during navigation, and addresses races when opening newly started sessions. It also improves background-session retention and worktree cleanup behaviour.→Better Auth 1.7.3 restores schema compatibility and tightens authentication defaultsThe release restores compatibility with 1.6 account schemas and enables initialisation-time schema validation by default, rejecting authentication requests when detected mismatches remain. It also adds Cloudflare as a social provider, Have I Been Pwned password checks and fixes for OAuth callbacks, session handling and TOTP re-enrolment.→OpenClaw 2026.9.2 makes agent operations more recoverableThe release moves heavy transcript work away from the gateway event loop, expands settings that apply without restarts, adds GPT-6 Astra support and enables Swarm orchestration by default. It also changes cross-agent session visibility defaults, so shared Gateway deployments need to review their access settings.→LiteLLM 1.100.0 expands gateway integrations and security controlsThe release adds per-team New Relic trace routing, Grounding with Bing Search, native Vertex AI Interactions API support and Entra ID/OAuth authentication across Azure AI Foundry routes. It also introduces vector-store upload security controls and updates MCP, A2A, routing, cost and provider handling.→Microsoft APM 0.29.1 hardens agent package managementAPM 0.29.1 adds support for catalog-only marketplace plugins and improves diagnostics for malformed executable trust configuration. It also fixes staging cleanup, plugin hooks during reinstall, private cache authentication, deterministic lockfiles and MCP dependency resolution.→DOMPurify 3.4.15 strengthens XML clobbering defencesThe release adds better clobbering hardening when XML content is involved and includes several smaller hardening and edge-case improvements. It also updates dependencies where possible.→Bifrost HTTP transport receives a new unauthenticated RCE disclosureThe CVE describes unauthenticated plugin-path submission through POST /api/plugins when management authentication is disabled, allowing native code execution on affected dynamically linked builds. The published static Docker image is expected to be limited to SSRF, but operators should still upgrade or enable management authentication.→EvoAgentX publishes its first official self-evolving agent framework releaseThe initial release includes workflow orchestration, short- and long-term memory agents, RAG integration, external MCP, HTTP and CLI tools, and human-in-the-loop interaction. The project presents the framework as a modular substrate for building and evaluating self-evolving agent workflows.→NVIDIA agrees to acquire Hugging Face for $12.93 billionNVIDIA said it agreed to acquire Hugging Face for $12,930,300,000. It said Hugging Face will remain an open platform supporting open models, multiple clouds, accelerators and developer tooling.→Meta releases Muse Spark 1.3 for coding and agentic workflowsMeta made Muse Spark 1.3 available in Muse Code and the Meta Model API, describing improvements in long-horizon work, multitasking, constraint following and confirmation before consequential actions. Meta engineers reported approximately 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2 in coding comparisons, alongside stronger prompt-injection resistance.→GitHub details cost reductions in AI coding without reported quality lossGitHub described shipped Copilot changes including compression of repetitive command output, removal of unused line-number prefixes, shorter task prompts and direct delivery of completed background results. It reported approximately 5% lower offline benchmark inference cost, 3% lower average daily model-inference cost per user and a 2.3% reduction in token-related usage, with no material quality regression in its tests.→CodeQL 2.26.4 expands language and GitHub Actions security coverageGitHub rolled CodeQL 2.26.4 into code scanning on github.com with Go 1.27 support, more precise Rust data-flow locations, new Spring R2DBC SQL-injection sinks and expanded Java, Kotlin, JavaScript, TypeScript and Python taint modelling. The release also adds GitHub Actions detections for mutable references to reusable workflows and improves checks on workflow actor fields.→GitHub lets enterprises set Copilot's default model by teamGitHub made enterprise-managed default-model settings available for Copilot Business and Enterprise. Administrators can set the model used for new conversations and map different defaults to enterprise teams through configuration, including across the Copilot app, CLI and VS Code.→Codex CLI 0.153 adds plugin marketplaces and resumable agent sessionsCodex CLI 0.153.0 adds commands to list, install and remove plugins from remote marketplaces, along with experimental context management and new app-server thread metadata. It also improves recovery after app-server disconnects, preserves Guardian and MCP approval state across compaction, restarts and forks, and adds a fuller TUI history view.→VS Code 1.136 adds Agent Merge and multi-root agent workspacesVS Code 1.136 introduces Agent Merge in preview, allowing an agent to address review feedback, failed checks and conflicts and rerun workflows until a change is ready to merge. It also adds experimental multi-root workspaces for Copilot and Claude sessions and expands agent-session hierarchy, notifications and workspace resolution.→GitHub makes Copilot content exclusions available in the app and CLIGitHub made content exclusions generally available in the Copilot app and CLI. The clients respect enterprise, organisation and repository policies so excluded files are not supplied as agent context.→pnpm 12.3 adds trust-policy checks and native global-command shimspnpm 12.3 adds trust-policy flags to pnpm remove and pnpm update, verifying the lockfile against active policies. It also makes context-aware global commands such as node, deno, bun and pnpm-added shims native executables on every platform and improves local and tarball installs.→Hugging Face releases Funes, local memory for coding-agent tracesHugging Face introduced Funes, an open-source tool that indexes traces from Claude Code, Codex, pi and Hermes into a local memory or dataset. It can retrieve source passages instead of generated summaries and optionally share a dataset across machines or agents through Hugging Face.→JFrog adds AgentSecOps controls across the agentic software supply chainJFrog announced a suite connecting models, MCP servers, skills, plugins, prompts and other AI assets to its software-supply-chain platform. It includes semantic scanning, an agent-plugin registry, APM package management, project-scoped Agent Guard policies and network-layer routing through Artifactory.→CrowdStrike introduces Falcon Guardian for runtime control of AI agentsCrowdStrike described Falcon Guardian as a live inventory and runtime-control layer for approved and shadow agents, backed by endpoint telemetry. Its Codex integration is designed to detect compromised or unauthorised behaviour and enforce permitted actions across endpoint, SaaS, cloud and browser environments.→F5 and MuleSoft put inline AI guardrails inside Agent FabricThe integration is designed to centralise policy enforcement, real-time protection against malicious prompts and sensitive-data exposure, and auditability. F5 and MuleSoft position it as an inline control that can be added without replacing existing agent systems.→Traversal makes Incident Workers generally available and launches Alert WorkersTraversal announced general availability of Incident Workers, a public beta of Alert Workers and a Workers console for deploying and managing the fleet. Incident Workers investigate incidents end to end in Slack, Teams or the web console, while Alert Workers reason across signals, deduplicate noise and attach explanations to alerts.→AIR emerges with a vetting layer for AI-agent skills and add-onsAIR emerged from stealth with $50 million raised across two seed rounds and a platform that discovers and continuously vets agent skills, tools and other components. The company says its system can block unsafe external or software interactions and provide a marketplace of reviewed extensions.→GitHub Copilot CLI adds model fallback and proxy-bound Linux sandboxesThe v1.0.83-2 pre-release lets custom agents declare an ordered list of models and enforce a required model policy, and adds support for Claude Fable 5.1. It also restricts Linux sandbox network egress to a configured proxy, with additional host requirements for proxy mode.→Cline Desktop 0.0.22 imports resumable sessions from other coding agentsCline Desktop 0.0.22 can scan local stores and import selected Claude Code, Codex and OpenCode conversations as fully resumable Cline sessions. The release also folds scheduled runs into the sidebar and adds workflow, tool-call and marketplace fixes.→LangChain 1.4.0a3 adds native MCP adaptation and human resumptionsLangChain 1.4.0a3 introduces a `langchain.mcp` namespace that adapts MCP servers into LangChain tools from URLs, local scripts, in-process servers or prebuilt clients. Its elicitation interrupt mode surfaces server questions as LangGraph interrupts so a human can answer and resume the run, with optional tool-discovery caching.→Cloudflare workerd hardens Python workers and hibernating WebSocketsThe release adds automatic Python-worker flags, synchronous try-read/write helpers and fixes for Pyodide dynamic library loading. It also improves hibernating WebSocket write serialisation and actor-fetch retries, while expanding test coverage for streams and queuing strategies.→curl 8.22.0 ships security fixes and experimental HTTP Message Signaturescurl 8.22.0 includes security fixes across curl and libcurl, alongside 302 bug fixes and 525 commits. It adds experimental RFC 9421 HTTP Message Signatures, changes SPNEGO NTLM fallback behaviour and drops TLS-SRP, among other platform and protocol updates.→GitHub makes the billing organisation the authority for Copilot model accessModel availability is now determined only by the organisation paying for the user's Copilot usage. GitHub said users whose access comes entirely through an enterprise are unaffected.→Claude Code 2.1.252 fixes remote-session stalls and oversized failure outputVersion 2.1.252 fixes Remote Control sessions stalling after tools finish when the Claude service connection is degraded. It also prevents very large background-task failure messages from pushing conversations beyond the API request-size limit.→Codex CLI 0.152 adds MCP output caps and longer shell-command deadlinesThe release adds per-tool MCP output limits, package-style MCP server names and configurable shell-command timeouts beyond one hour. It also preserves approval context through compaction and rejects untrusted cloud-task URLs and redirects.→Cline Desktop 0.0.21 makes delegated-agent stops propagateStopping a session now propagates aborts to child agents and teammates, with cancelled teammate tasks persisted as cancelled. The release also refreshes live provider models, distinguishes authentication failures from outages, restores Langfuse tracing in production builds and changes unpinned defaults across about 36 providers.→Next.js 16.4 canary.13 improves Turbopack HMR recoveryThe release fixes Turbopack HMR recovery after dropped connections and preserves original search parameters for private-cache pages. It also adds adoption-skill evaluations and documentation for the create-next-app Cache Components prompt and flag.→Cloudflare workerd adds file-system writable streams and compression streamsworkerd v1.20260901.1 implements FileSystemWritableFileStream in TypeScript and adds encoding-stream and compression-stream suites with migrated tests. It also catches leaked SQLite handles during initialisation failure.→Apodex 1.1 pairs an open-weight model with verifiable executionApodex 1.1 combines multi-agent collaboration, tool execution, failure recovery and source-linked verification in a persistent workbench that can process code and other artefacts. The release also includes open weights for Apodex 1.1 Mini and the open-source FrontierAgent framework; its capability comparisons are company-published.→Cycode releases agentic code scanning and attack-path chainingCycode said its Agentic Code Scanning combines rule-based and agentic analysis, while its benchmark testing found two authorisation CVEs that rule engines could not express. Its new Attack Chaining capability links related findings into multi-step exploit paths.→Reco releases Browser Guard for browser-based AI and coding-agent actionsBrowser Guard inspects prompts and agent activity, correlating them with user, data, destination and business context. Reco said the product can detect, guide, block or stop activity involving AI assistants, copilots, coding agents and autonomous workflows.→Ping Identity adds runtime identity controls for Claude and coding agentsEnterprise Personal Agent Access, delivered through PingOne Privilege, associates agent sessions with the user and device behind them and applies allow, deny, logging and approval policies at runtime. Ping said the controls cover desktop assistants such as Claude and coding agents such as Claude Code, including commit attribution without long-lived credentials.→Broadcom launches TrueSource for verifiably built open-source softwareTrueSource combines Spring Enterprise with clean-room builds across Java, Python and Node.js, hardened container images, and trusted data services for PostgreSQL, RabbitMQ, MySQL and Valkey. Its tooling scans repositories, assesses release blast radius, proposes lower-risk remediation pull requests and provides security dashboards.→Broadcom adds runtime identity, network defence and observability for agentic AIAgentMinder binds agent authority to declared missions, intents, tools and resources, enforcing least-privilege policy at each tool invocation. VMware vDefend and Avi add MCP and LLM discovery, shadow-AI monitoring, anomalous-traffic detection, malicious-execution blocking and OpenTelemetry auditability.→JetBrains confirms Cadence server-backup access in incident updateJetBrains said Cadence, a hosted service for running PyCharm projects on cloud compute, was accessed through a compromised environment. The August 31 update confirmed access to a 2024 Cadence server backup while stating that no additional compromised resources or data had yet been identified.→Broadcom introduces VMware Private AI Cloud for private inference and agentsThe platform combines VMware Cloud Foundation with heterogeneous accelerator support, NVMe memory tiering, storage deduplication, token monitoring, multi-tenant model sharing and GPU observability. Broadcom also says customers can access more than 150 validated open and commercial models through integrated deployment and operations services.→Tealium ships governed Configuration MCP and reversible platform APIsThe Configuration MCP lets compatible agents create and modify audiences, attributes, enrichments and activation workflows, with changes held as drafts until review. The Developer Portal adds authenticated, versioned, validated and reversible workflows with access controls and rollback support.→Data Dynamics makes governed AI data access generally availableEnterprise 2.0.5 adds bring-your-own-model deployment inside an organisation’s trust boundary and content-aware, time-bound entitlements for people, models and agents. It also provides attestable compliance evidence and data-lifecycle controls across cloud, on-premises, sovereign and air-gapped deployments.→QA Cafe automates long-duration stability testing for broadband devicesThe new testing workflow repeatedly exercises broadband and Wi-Fi devices while monitoring CPU and memory through TR-181 over TR-069 or USP. It correlates resource trends with pass/fail results, logs and packet captures to expose leaks, exhaustion, crashes and gradual performance degradation.→Forward opens its network digital twin and AI workflow to wider teamsThe trial lets network, security and cloud teams ask plain-language questions against a mathematically accurate model and receive evidence-based answers. It can also identify the assets affected by a CVE and provide deterministic evidence about how proposed changes affect production networks across major clouds.→Think launches hosted heterogeneous AI compute with Think GridEach initial SuperNode combines four NVIDIA PRO 6000 Blackwell accelerators with 384GB of pooled VRAM, networking, storage, cooling and support. The service is available to Saudi government and enterprise customers from Riyadh and is being rolled out globally for inference, training and agent workloads.→HUMAIN and MinIO partner on a sovereign AI data fabricThe companies announced a strategic engineering and go-to-market partnership covering HUMAIN Fabric, HUMAIN Brain and HUMAIN One. MinIO will lead the architecture, design and development of HUMAIN Fabric, an AI-native data platform intended to support enterprise and sovereign AI deployments beginning in Saudi Arabia.→T3 Code nightly unifies agent activity and composer stateT3 Code version 0.0.37-nightly.20260830.1225 unifies activity logs with composer banners in the web interface. Its mobile changes reduce development-client reloads and Metro startup cost.→Next.js 16.4 canary improves codemods and SST key orderingNext.js 16.4.0-canary.12 stores SST keys in key order when hashes are omitted and honours non-interactive mode in upgrade codemod prompts. The release also extends codemod transformation to .mjs files and removes an unnecessary CI workflow.→quic-go 0.62 brings RFC 9218 priorities to HTTP/3quic-go 0.62 adds RFC 9218 stream priorities, including SetPriority APIs and HTTP/3 handling for Priority headers and PRIORITY_UPDATE frames. It also supports reliable stream resets across newer and legacy drafts and raises the minimum Go version to 1.26.→GitHub Copilot CLI 1.0.82 tightens worktree and approval flowsVersion 1.0.82 fixes a failure triggered when users typed during /worktree or /move preparation. It also expands plan approval cards and reports specific authentication failures instead of showing only a generic login prompt.→Alpine.js 3.17 adds deferred tree initialisation and a CSP fixAlpine.js 3.17 adds Alpine.deferInit(), allowing a component tree to remain suspended until an asynchronous prerequisite settles. The release also fixes Unicode escapes in Content Security Policy string literals.→Angular ESLint 22.2 adds configurable explicit OnPush handlingAngular ESLint 22.2 adds an allowExplicitOnPush option that can automatically remove explicit OnPush declarations when a project disallows them. It also reports Angular major-version mismatches at runtime and adds template rules covering duplicate metadata arrays, ARIA validity and static string properties.→jscpd 5.1 turns duplication baselines into CI gatesjscpd 5.1 introduces content-hash baselines, baseline updates and a fail-on-new-clones mode so CI can reject newly introduced duplication without blocking existing debt. It also adds ephemeral baselines from Git references, OpenMetrics and CodeClimate or GitLab Code Quality reporters, plus Windows ARM support and signed releases with SLSA provenance.→Plasmo 0.89 adds browser-specific manifest overridesPlasmo 0.89 adds an overrides key for browser-specific manifest configuration and logs the selected target during plasmo build and plasmo dev. The release also documents zip output and CLI argument values more clearly.→Javy 6.0 changes its plugin API for JavaScript-to-WebAssembly buildsJavy 6.0 removes eval_bytecode from the default plugin and update namespace, moves plugin validation into code generation and changes the configuration schema API. It also adds a code-generation option for including source and documents a WASI Preview 2 plugin direction.→React Hook Form 7.87 adds opaque TypeScript leaves and shouldTouch validationReact Hook Form 7.87 adds a shouldTouch option to trigger() and an OpaqueTypes registry for leaf types that should not be recursively analysed. The release also fixes controller change handling, hidden Activity subtree watching, resetField validity and React Server declaration resolution in Next.js builds.→Claude Code adds restricted execution and cross-session messagingThe release adds --restricted mode, which removes command-running tools and WebFetch unless explicitly allowed, confines file tools to the working directory, refuses bypassPermissions, and ignores local settings files. It also adds cross-session SendMessage and ListAgents support on Bedrock, Vertex, and Foundry, alongside self-hosted runner labels and prompt-cache fixes.→Claude Code adds model-switch hooks and remote subagent visibilityThe release adds PreModelSwitch and PostModelSwitch hooks, plus live streaming of foreground subagent tool calls and results to Remote Control clients. It also exposes spend-limit and prompt-cache data and fixes symlink races, plugin path traversal, unsafe workflow paths, and symlink-based search-rule bypasses.→OpenCode adds Azure Entra sign-in and fixes replayable Bedrock reasoningOpenCode v1.18.24 stops Bedrock reasoning responses from being cached as unreplayable empty messages and lets Azure providers authenticate through Microsoft Entra ID via the Azure CLI. It also reads supported v2 configuration fields, easing mixed-version deployments.→Cline Desktop expands to Windows and improves long-running agent workflowsThe release ships a code-signed x64 Windows installer with auto-updates and adds full indexed session search, inline rendering for tool images, and a GitHub integration step. It also improves scheduled-task persistence, checkpoint restoration, background processes, and remote MCP handling.→SkyWalking 11.0 separates Horizon UI from OAP and adds live rule debuggingSkyWalking 11.0.0 makes Horizon UI an independently deployed component rather than a bundled legacy web UI and exposes a dedicated administration port for management APIs. The release adds TLS certificate hot reload for OAP HTTP and REST servers, runtime rule hot-update, live DSL debugging, and trace-tail sampling.→Cloudflare opens BotBase controls for bot and agent operatorsThe update adds a dashboard home with a bot directory, submission history, status explanations, and controls to edit or cancel submissions. It also expands classification around behaviour, content use, and identity while leaving site owners in control of how traffic is handled.→GitHub Agentic Workflows adds grader audits and hardens MCP executionThe release adds daily grader-audit workflows, cross-run audit clustering, and pattern analysis to its logs and audit commands. It also tightens label permissions, MCP gateway readiness, fork pull-request safe-output handling, scanner argument handling, and dependency security gates.→Huawei Cloud moves CodeArts Agent to general availability in Asia PacificThe Basic and Professional editions are now generally available and combine project-level code understanding with autonomous development workflows. CodeArts Agent includes 16 specialised agents spanning requirements, architecture, coding, testing, issue resolution, and code review, plus more than 30 reusable engineering skills and IDE, plugin, CLI, and TUI access.→Optimizely adds WebMCP tools to Spire commerce storefrontsThe release adds a WebMCP layer to Spire storefronts for reading session and cart context, searching products with account-specific pricing, and adding, updating, or removing cart lines. It also improves server-side rendering resource use and adds storefront error timestamps and reference IDs for faster troubleshooting.→Vercel adds dashboard deployment for eve agentsThe builder scaffolds an eve agent, pushes the code to a private Git repository, and creates a Vercel project. Developers can select models through AI Gateway and configure a Next.js web chat or Slack interface with built-in Linear, Notion, and custom MCP connections.→SonarQube Hunter Agent adds reasoning-based security analysisSonar launched Hunter Agent for SonarQube Cloud to analyse broken access control, business-logic, authentication and session-management flaws. The agent traces code, data and identity flows across a codebase and verifies candidate findings before presenting them in the existing Sonar workflow.→Harness launches an agent-ready code repository with AI code reviewHarness launched Agent-Ready Harness Code Repository and AI Code Review for teams adopting coding agents. The release combines agent-scoped RBAC and OPA permissions, MCP and CLI access, risk-grouped review, mandatory checks and one-click remediation with migration paths from major repository platforms.→Operant launches an intent-aware firewall for coding and enterprise agentsOperant AI launched Semantic Firewall to allow, block or redact prompts, tool calls, commands and data payloads in real time. Its Code Intent Guard is aimed at package installation, MCP and tool setup, shell commands, prompt injection, privilege escalation and hidden directives.→Redwood makes its orchestration platform agent-ready with MCP and AWS connectivityRunMyJobs 2026.3 added a generally available MCP server with more than 50 production-ready tools, native AWS S3, EventBridge and Glue connectivity, and support for agent-to-agent workflows. Redwood also introduced Agent Studio, Workflow Builder and an Operations Agent as technology previews with audit and governance controls.→Workday unveils a developer agent and passport for enterprise AI agentsWorkday unveiled Developer Agent, which lets developers create AI applications and agents through natural-language interaction with agentic development tools. It also announced Agent Passport, which tests and verifies Workday-built and third-party agents before production and monitors them afterwards.→SerpApi adds Markdown responses for lower-token web search payloadsSerpApi added optional Markdown responses through query parameters, endpoints and headers across its API catalogue. The company reports an average token reduction of about 50%, with larger reductions for some APIs, while keeping the existing integrations available.→PaperCut releases an emergency patch after confirmed exploitationPaperCut's security bulletin recorded confirmed exploitation and customer incidents affecting all NG and MF versions, then announced an emergency patch for versions 25 and 26 at 02:10 AEST. The company advised restricting public access to affected servers while work continued on the version 24 patch.→Ars Technica reports a new web-based supply-chain risk for coding agentsArs Technica reported 227 installation commands across more than 100 websites that pointed to unowned executable content in files such as llms.txt and llms-full.txt. The report said Claude, Codex and Hermes agents installed the code, exposing a new route from web documentation to developer workstations and corporate networks.→Echo acquires Minimus assets to expand hardened open-source software coverageEcho acquired Minimus technology and data to broaden its hardened-artifact operating-system coverage and vulnerability-scanner integrations. Echo said the combined offering will provide distribution-agnostic hardened open-source artefacts, with agents used to investigate vulnerabilities and develop and validate patches.→Above Security and Forscie publish an open taxonomy for AI-agent insider threatsAbove Security and Forscie launched the Synthetic Insider Threat Matrix, an open taxonomy modelled on MITRE ATT&CK. Its 166 knowledge objects cover behaviours including unauthorised data access, autonomous exfiltration, privilege misuse and shadow AI activity.→Codex CLI 0.150 makes cross-task coordination and trust controls first-classCodex CLI 0.150 lets users reference other tasks with @ mentions and ask agents to read, create or message tasks from the terminal; it also adds a copy picker, automatic task titles, clickable Markdown links and interrupt hooks. The release says untrusted projects no longer provide project-level AGENTS.md instructions, managed deny-read rules persist after permission changes, and app-server diagnostics improve credential redaction.→VS Code 1.135 turns the Agents window into a cross-agent workspaceVS Code 1.135 can show recent Copilot or Claude agent sessions created in other applications and let users continue them in VS Code. It also introduces experimental Rubber Duck second opinions, a default single-pane Agents layout and per-model input, cached-input and output token details.→Kubernetes 1.37 graduates major scheduling, policy and storage capabilitiesKubernetes 1.37 ships 67 enhancements, including stable StorageVersionMigration, metrics.k8s.io and KYAML, plus beta HPA scale-to-zero, manifest-based admission control and native gang scheduling. It also adds alpha Pod-level checkpoint and restore and CompositePodGroup APIs, while resilient watchcache initialisation reduces startup and recovery traffic spikes against etcd.→Anthropic updates Claude Code and its TypeScript Agent SDKClaude Code v2.1.246 adds an Auto mode tab to /permissions and fixes background-session startup, MCP interruption reporting, plugin handling, large-session safety checks and third-party gateway credential handling. TypeScript Agent SDK v0.3.246 adds user_message_uuid correlation, costBasis and managed modelPricing metadata, plus perTaskStopAffordance so interrupt() can stop the current turn without stopping background agents and workflows.→Node.js 26.8 adds crypto, diagnostics and SQLite APIsNode.js 26.8 enables SIV and GCM-SIV modes in Cipher and Decipher, marks TracingChannel stable, improves histogram and net.BlockList behaviour, and adds non-throwing MIMEType.parse. It also adds SQLite StatementSync close and dispose methods, basic REPL syntax highlighting and ZipFile, ZipEntry and ZipBuffer support, alongside certificate and Web Crypto fixes.→pnpm 12 makes dependency resolution and runtime selection more deterministicpnpm 12 canonicalises known-host Git dependencies to HTTPS identities, makes peer-dependency cycle breaking deterministic and fails or warns on unrecognised workspace settings instead of silently ignoring them. It adds project-aware shims for pinned runtimes, can provision other package managers and introduces an opt-in signed remote side-effects cache.→Hono 4.13.5 patches three web-framework security issuesHono 4.13.5 fixes query parsing that could read parameters after a URL fragment, creating cache-key and proxy interpretation differentials. It also closes a toSSG path-normalisation gap that could write outside the output directory and bounds dot-notation expansion in parseBody() to prevent memory exhaustion.→AI-Infra-Guard 4.6 adds LLM API-poisoning checks and agent red-team coverageAI-Infra-Guard 4.6 adds an API Checker security-audit module with web-proxy integration, a detector for LLM API poisoning, a v5 Agent-Scan mutation-engine refactor and a DeepSeek Harness prompt-injection assessment. It also tightens YAML rule validation, migrates MCP clients to SDK 2.0 and fixes parser, filename and JSON-serialisation failure modes.→GitHub Copilot SDK makes empty-mode skills deny-by-defaultCopilot SDK v1.0.13-preview.1 makes ClientMode.Empty apply deny-by-default isolation to built-in skills across all six SDKs. Hosts can re-enable selected skills through includedBuiltinSkills, while sessions created in Empty mode otherwise start without runtime-bundled skills.→Vercel AI SDK adds a Z.AI provider with GLM-5.3 Flash supportThe @ai-sdk/zai 2.0.0 release adds a Z.AI provider with GLM chat completions, streaming, reasoning, tool use and multimodal inputs. It also backports GLM-5.3-Flash model support to the Z.AI provider and AI Gateway.→TRACE proposes a portable evidence standard for AI-agent runtimesOPAQUE said TRACE was developed with AMD, Intel, Microsoft and the Technology Innovation Institute and contributed to the Linux Foundation. The specification combines attestation, provenance and workload-identity standards to record software, policies, data classifications and tools used by an AI workload.→AC2 launches an open protocol for user-approved agent actionsAC2 uses encrypted WebRTC connections and lets an agent request actions such as payments, API authorisations or Git commits for approval in the user's wallet or application. The foundation said private keys remain under user control while the open-source specification and reference implementation are available on GitHub.→Nucleus Helix combines vulnerability intelligence with agentic exposure workflowsNucleus said Helix adds passive early-warning discovery before scanner signatures are available, a natural-language AI agent and expanded data-collection and remediation guidance. The platform is grounded in Nucleus's asset, software, ownership and exposure data.→A malicious webpage can poison the local model behind NVIDIA NemoClawThe researchers said a malicious webpage can use DNS rebinding to reach the local Ollama API without credentials. They reported that an attacker can change the model's chat template, allowing hidden instructions to persist across later agent conversations.→Anjuna brings confidential containers to on-premises AMD SEV deploymentsThe release adds AMD SEV and Secure Nested Paging support for confidential containers without application refactoring. Anjuna also described unified policy management, remote attestation and air-gapped deployment across cloud and on-premises environments.→CogniCrypt applies concolic execution and LLM-guided analysis to AI-generated malwareThe company said CogniCrypt combines concolic execution, LLM-guided analysis and deep learning to analyse unfamiliar software. Quandary Peak is incorporating it into M&A assessments and described the underlying work as accepted at the CRIS 2026 conference.→OpenCode 1.18.23 repairs Cloudflare AI Gateway routing and session handlingVersion 1.18.23 fixes third-party provider routing through Cloudflare AI Gateway and converts dotted Anthropic model IDs to the dashed form expected by Anthropic. It also fixes parent-session headers for session-aware providers and GitHub authentication using immutable OIDC subject tokens.→Cline Desktop 0.0.17 consolidates agent controls and caps event-log growthCline moved plugins, MCP servers, skills, rules, hooks and tools into one Customise hub and redesigned provider management. The release also fixes codebase-search crashes on files containing a single enormous line and prevents the hub event log from growing until it fills the disk.→EPAM and Wiz pair cloud-risk findings with engineering-led remediationEPAM said it will use Wiz's Security Graph and AI-APP across code, cloud, runtime, AI applications, SaaS and custom-hosted environments. The partnership adds EPAM engineering and White Hat offensive-security capabilities to help validate and remediate exploitable paths.→Google Antigravity CLI 1.1.20 improves workspace approvals and repository handlingThe release adds workspace-scoped read auto-approvals, indexes empty directories for path completion and skips recursive worktree scans in submodules. It also makes benign tool errors and permission denials non-fatal in print mode and preserves malformed settings instead of discarding them.→OpenAI ships Codex CLI 0.149.1Codex CLI 0.149.1 was released on 24 August. GitHub links the 0.149.0-to-0.149.1 changelog, although the release page does not enumerate the changes.→Next.js canary 16.4.0-canary.5 advances PPR caching and TurbopackThe release records a cache-lifetime change for blocking PPR routes and new documentation for Pages Router catchError. It also includes Turbopack emit and collect work, custom-server response-header documentation, dependency deduplication and a cold-path performance split in TurboMalloc accounting.→Hono 4.13.4 fixes request, routing and client edge casesHono 4.13.4 fixes request-parameter handling on unmatched requests and wildcard middleware association in the regexp router. It also corrects JSX ref cleanup, ETag parsing, undefined header and cookie values, form-data cloning, client query serialisation and media-type matching.→Vite+ 0.3.0 moves installs to standard platform directoriesFresh installs now use split XDG directories on Unix and standard local application-data folders on Windows, while existing single-root installations remain in place. The release adds tsup-to-vp pack migration, forwards supported commands to Bun 1.4, cuts the published package size by 51 percent and removes upgrade-check delays.→Microsoft moves WebView2 Runtime to a two-week release cadenceVersion 151 is the final four-week release, with version 152 planned for late August and version 153 planned for the week of 10 September. The SDK remains on a fixed monthly cadence and ships as needed, while evergreen apps update automatically and fixed-version apps require more frequent testing and updates.→AccuKnox launches AgentZ for building, running and governing AI agentsAgentZ provides organisations, workspaces, agents, workflows and sandbox objects, with role-based access, runtime credential management, triggers, skills and governance controls. AccuKnox positions it as a platform for deploying agents across teams and business workflows rather than as a model or single-purpose assistant.→Okta makes Agent SSO generally available for AI-agent identityThe service is designed to replace static API keys and repeated user consent with central policy, short-lived tokens and delegated access. Okta says Agent SSO can govern agents as they call applications and MCP servers, giving administrators a control point for identity, permissions and revocation.→Zide opens public beta of a native workspace for code, Git, issues and CIThe public beta is available for macOS, Windows and Linux and supports GitHub, GitLab and Bitbucket. Zide Assist can use Claude, Codex and Gemini, while the application is built with Tauri and Rust rather than as a browser wrapper.→Braid demonstrates local, cross-platform AI state exchangeThe release demonstrates state exchange across Windows, macOS and Linux, with local developer diagnostics for inspecting exchanged AI state. Intersignal says the current system works over LAN or Wi-Fi and that managed NAT traversal is planned, while the core experience remains free.→create-vite 9.2.0 adds nub package-manager supportcreate-vite 9.2.0 adds nub package-manager support and documents experimental React Compiler use in generated projects. It also updates eslint-react links and related development dependencies.→Cline SDK v0.0.78 adds durable Hub event replayQueued runs now survive Hub upgrades, reconnecting clients replay missed events, and event IDs prevent duplicate processing. The release also adds session and client identity to Langfuse traces and refreshes model defaults and pricing.→Cline CLI v3.0.57 adds drain-aware Hub upgradesThe new commands drain a Hub, wait for active work to finish, restart the current build, and lift the drain if an upgrade is aborted. Sessions replay missed events after reconnecting, while custom OpenAI-compatible models receive corrected tool-calling behaviour.→Cline v4.1.14 refreshes its supported model catalogueThe release adds catalogue entries spanning Anthropic, xAI, DeepSeek, Qwen, Google, Meta, NVIDIA, and other providers. It also fixes interactive task completion telemetry so a completion is reported exactly once.→Google Antigravity CLI 1.1.19 improves remote control and terminal accessibilityThe remote-control command no longer fails when its preferred port is occupied; it now requests a free operating-system port. New environment variables can hide the logo or disable escape-sequence optimisations for screen readers, recordings, and constrained terminals.→LiteLLM v1.98.0 adds gateway observability and control featuresThe gateway can now attach opt-in session and trace identifiers to JSON logs, send per-deployment SSE heartbeats, and enforce configurable output-token limits by key, team, or model. The release also exposes routing groups as virtual models and improves MCP timeouts, tracing, and provider cost accounting.→Vercel AI SDK fixes Deepgram transcription and speech integrationOptions including keyterms, paragraphs, intents, sentiment, and replacement are now sent to Deepgram instead of being silently ignored, while diarisation is now opt-in rather than enabled by default. The provider also composes voice identifiers, passes speech speed, exposes response metadata, and parses Deepgram's structured error responses.→Next.js 16.4.0-canary.2 introduces backend-storage construction optionsThe canary release adds an options struct to the backend-storage construction path. The published release notes describe this as the release change and do not announce a broader stable feature set.→redb v2.6.3 fixes range-bound correctnessThe release corrects range(), extract_from_if(), and retain_in() when a range starts after it ends and the keys are separated across internal B-tree pages. Previously, those cases could be treated as an unbounded start range.→PocketBase v0.40.0 hardens defaults and backend operationsThe release enables a same-origin Cross-Origin-Opener-Policy header by default, adds log deletion, caps log payload sizes, and avoids transaction locking during backups. It also enables SQLite defensive mode, updates modernc SQLite, raises the minimum Go version to 1.27, and migrates to encoding/json/v2.→Cline Desktop v0.0.16 makes Hub restarts recoverableThe release drains a Hub before restart, refuses new work, and replays missed messages so agent handoffs survive upgrades. It also refreshes model catalogues and pricing and honours server-side feature flags.→Cloudflare workerd adds dynamic WebAssembly modules and TypeScript stream RPC supportworkerd v1.20260822.1 adds `WebAssembly.Module` support to the dynamic worker loader and implements jsrpc serialisation and deserialisation for TypeScript web streams. It also upgrades SQLite to 3.53.4 and fixes an alarms-in-facets error.→Gemini CLI nightly hardens macOS agent sandbox boundariesThe v0.56.0 nightly 20260822 build isolates Docker and container-runtime sockets and binaries in macOS Seatbelt. The change is a pre-release sandbox fix rather than a stable-channel feature.→GitHub Agentic Workflows tightens safe outputs and reproducibilitygh-aw v0.87.4 makes compilation fail when safe-output step-output tokens cannot resolve, adds pinned Agent Plugins, and supports per-engine model overrides and custom providers. It also stops `GH_TOKEN` persisting in repo-memory clone `.git/config` and derives conclusion-job write permission from resolved safe-output configuration.→Cline 4.1.12 extends enterprise MCP controls across its agent stackCline 4.1.12 hides MCP marketplace entries when remote configuration disables the marketplace or restricts it to an `allowedMCPServers` allowlist. It also restores tool calling for custom OpenAI-compatible models whose stored capability list was empty.→Next.js 16.4 canary advances Turbopack watcher and navigation validationNext.js v16.4.0-canary.1 changes Turbopack’s filesystem watcher to avoid cloning paths already held in its map and adds instant validation for `unstable_navigation()`. It also improves graceful-fs tracing, missing-root-layout errors and Partial Prefetching adoption checks.→Aspire 13.5.2 removes an unused Windows CLI binary blocking WinGet publicationAspire 13.5.2 removes the unused `hex1bpty.exe` from Windows CLI archives because Aspire never executes it. The patch lets the release pass WinGet’s executable and malware validation while leaving Unix native assets unchanged.→syslog-ng 4.9 adds native HTTP statistics exporterssyslog-ng 4.9.0 adds `stats-exporter()` and `stats-exporter-dont-log()` sources that expose `syslog-ng-ctl` statistics or queries to HTTP scrapers, with rate, format and single-instance controls. It also adds RFC6587 auto-detection for octet-count framing and inotify-based wildcard-file change detection.→React Hook Form 7.86 adds type-safe errors and reduces form-state overheadReact Hook Form 7.86.0 adds a type-safe `getErrors` method and fixes stale watch, unregister, validation-state and `setCustomValidity` behaviour. It also improves `createFormControl` and `onChange`-heavy workloads while avoiding unnecessary value cloning.→pybind11 3.0.1 stabilises sub-interpreters and difficult C++ buildspybind11 3.0.1 fixes enum-pointer caster compilation, reduces template depth in `make_index_sequence`, and resolves sub-interpreter segfault and re-import issues. It also adds C++20 support for older C++ runtimes and addresses several compiler warnings.→PostgREST 16.2 improves JWT-role configuration compatibilityPostgREST 16.2 makes `jwt-role-claim-key` configuration backwards-compatible while deprecating the old pre-v16 JSPath syntax. Deployments using the old form receive a log warning and can follow the documented migration path.→Claude Code v2.1.238 hardens plugin credentials and self-hosted agent operationsClaude Code 2.1.238 added a headersHelper path for marketplace and MCP requests, with trust-gated execution and no inherited credential environment. It also added self-hosted-runner proxy and shutdown controls, improved Remote Control resilience, and fixed memory growth in long interactive sessions.→Codex CLI 0.149 turns parallel agent sessions into a managed workspaceCodex CLI 0.149.0 added an interactive codex agents dashboard for finding, starting, renaming and stopping tasks, plus codex queue for messaging existing local or remote sessions. The release also adds working-directory commands, broader diagnostics and SDK controls for reasoning effort and CLI configuration.→GitHub's Copilot Cloud Agent status visibility degraded for hoursGitHub's status page said Copilot Cloud Agent tasks were still completing, but newly started tasks did not show ongoing progress and session output was delayed by about an hour. GitHub reported gradual recovery during the window and marked the incident resolved at 00:37 UTC on 21 August.→Pydantic AI v2.33 repairs the Anthropic SDK 1.0 compatibility breakPydantic AI 2.33.0 requires and supports anthropic>=1.0.0 after fresh pydantic-ai[anthropic] installs could otherwise resolve an incompatible SDK and fail at runtime. Applications that pass a custom client to AnthropicProvider must now use httpx2.AsyncClient.→OpenCode v1.18.20 makes subagent failures resumable and provider retries explicitOpenCode 1.18.20 now surfaces failed subagent calls with resumable task IDs instead of returning empty results, and retries several network, capacity and temporary-provider failures. It also preserves provider-specific token settings and handles permission requests raised by subagents during opencode run.→ESLint v10.9 tightens autofix safety for JavaScript codebasesESLint 10.9.0 adds checks for conditional loop expressions and numeric precision underflow, while fixing unsafe no-var and prefer-template autofixes. The release also updates shareable-config documentation and its CodeQL and CI dependencies.→Cline CLI v3.0.56 carries skills, images and hook state into agent sessionsCline CLI 3.0.56 lets image-capable models return media through the TUI, HTML exports and ACP clients, and changes skill slash commands to load through the skills tool without duplicating their instructions. It also restores provider-tool visibility, hook output and cancellation behaviour, and fixes remote schedule authorisation.→Claude Agent SDK adds explicit subagent depth and refusal lifecycle eventsClaude Agent SDK TypeScript 0.3.238 adds is_backgrounded and spawn_depth to subagent task events, plus a terminal refused state for cross-session messages rejected by policy. It also fixes hook reinitialisation and exposes more reliable branch and prompt-suggestion events.→OpenHands v1.15 makes local agents and automation easier to inspectOpenHands 1.15.0 adds an LLM provider-connections UI for local agent servers, automation catalogue entries that ship script bundles, and a dashboard for managing automations. It also adds workspace-path visibility, a unified commits drawer and fixes silent agent-profile downgrades and workspace-folder pagination.→Cline SDK v0.0.77 scopes scheduled-task tools to capable clientsCline SDK 0.0.77 scopes durable todo and one-time or recurring scheduling tools to clients that can service them, with hosts declaring their client type. The release centralises tool-catalog resolution and prevents CLI and editor clients from exposing actions they cannot execute.→Bun 1.4 ships the Rust rewrite and broader full-stack toolingBun says v1.4 adds 1,517 Node tests, fixes 2,900 issues, cuts idle CPU use by five times, reduces memory use by up to 35 percent and starts up to 50 percent faster on Linux. The release also adds Bun.Image, Bun.WebView, Bun.markdown, Bun.cron, Bun.Terminal, parallel test execution and package-manager audit and repair commands.→Kubernetes 1.37 reaches release candidate statusThe Kubernetes project has published v1.37.0-rc.1 as a pre-release for final compatibility testing. The release schedule targets Kubernetes 1.37.0 for 26 August 2026.→TypeScript 7.0.2 publishes the native compiler releaseTypeScript 7's native toolchain uses a Go-based compiler and parallelised checking, with Microsoft reporting substantial build and language-server speedups on tested codebases. The 7.0 line also changes several defaults and does not yet provide a stable programmatic API for some framework integrations.→Spring Boot 4.2.0-M1 adds AMQP 1.0 and image-based Buildpacks cachesSpring Boot 4.2.0-M1 adds generic AMQP 1.0 and RabbitMQ AMQP 1.0 support, and lets Cloud Native Buildpacks publish and restore build caches through an image registry. Its release notes also change forwarded-header configuration and stop test MeterRegistry beans entering Micrometer's global registry by default.→Slack introduces shared code channels for team-visible coding agentsSlack Code lets a user tag an agent from a conversation and create a project code channel for the task. The Register reports live previews, searchable session history, human pause and redirect controls, and integrations involving Anthropic, Cognition, GitHub, OpenAI and Vercel.→Rust 1.98 adds new lints, targets and standard-library APIsThe release adds lints for invalid runtime symbol definitions, suspicious runtime symbol definitions and c_void return types. It also promotes several embedded targets and stabilises APIs including str::substr_range, algebraic floating-point operations and UTF-16 conversion helpers.→Cloudflare workerd expands Workers image, browser and workflow primitivesCloudflare's v1.20260820.1 release raises the SQLite row limit to 4 MiB and adds Images binding features including signedUrl and createDirectUpload. It also exposes Workflow instance location hints, adds Browser Run accessibility-tree types and permits current-process debug-port access.→Cline Desktop adds local-to-cloud agent handoffThe beta moves a conversation, attached images and an optional follow-up command into a Cline Cloud workspace that continues after the desktop app closes. It checks that the repository, branch and commit are pushed before handoff, adds Cloud to the existing Local/Remote menu, and restores drafts and attachments after interrupted transfers.→OpenCode adds native Cloudflare AI Gateway passthroughsThe release adds native OpenAI and Anthropic passthroughs for Cloudflare AI Gateway models and aligns Codex rate limits more closely with ChatGPT subscription limits. It also fixes malformed pricing handling, forwards ChatGPT workspace compute residency to Codex requests, enables web search for OpenCode Go and preserves v1 database compatibility.→Claude Code v2.1.237 fixes gateway prompt caching and adds Concise outputClaude Code v2.1.237 fixes prompt caching for sessions that use an LLM gateway or custom base URL. It also adds a selectable Concise style that leads with results and suppresses preamble and narration while preserving the work itself.→GitHub brings enterprise-managed Copilot settings to JetBrainsAdministrators can govern plugin and marketplace use, MCP server access, OpenTelemetry configuration, and permission modes. The controls can disable bypass or autopilot behaviour and restrict which plugins, marketplaces, and MCP servers developers may use.→GitHub adds organisation-level Code Quality trendsThe dashboard now plots open findings over 7-, 14-, or 30-day periods and ranks repositories by changes. It respects existing filters and is available to eligible GitHub Enterprise Cloud and Team organisations using Code Quality, but not GitHub Enterprise Server.→GitHub expands credential revocation by token typeOrganisations can now revoke or deauthorise credentials by token type and user, covering personal access tokens, SSH keys, OAuth app tokens, and GitHub App user tokens. The update adds organisation-level UI and API parity together with audit-log and email notifications.→Visual Studio Code 1.134 makes agent sessions easier to manageThe release lets the same agent session connect to multiple VS Code windows through the Agent Host Protocol and aligns its Copilot agent with Copilot CLI and other Copilot products. It also adds side-by-side chat groups, a prompt timeline with file-change counts, full-text conversation search, and local HTML preview in the integrated browser.→Gutenberg 23.8 adds shareable revisions and faster List ViewThe release adds shareable visual-revision URLs, a code-diff view, mention email notifications, and major List View performance work; selecting all 1,000 paragraphs fell from 16.8 seconds to 0.4 seconds in the release notes' test. It also adds inner-block templates to registerBlockType(), React 19 import support, DataViews time and aspect-ratio fields, and public Calendar and RangeCalendar components.→Claude Code adds default-model control and cross-session idle notificationsThe release adds ANTHROPIC_DEFAULT_MODEL for new sessions and a one-shot notify_when_idle option for cross-session SendMessage on macOS and Linux. It also tightens macOS sandbox deny rules, improves auto-mode command review, and adds screen-reader support for transcript events in VS Code.→Terraform 1.16.0-rc2 expands plan/apply data and lifecycle actionsThe release candidate adds provider-private data preservation across plan and apply, a store block in terraform_data for ephemeral and sensitive values, import blocks inside modules, and resource-action failure modes. It also adds JSON output for state show and workspace list, Mermaid output for graph, and destroy=false lifecycle support.→Codex CLI 0.148 adds session forking, MCP hooks and cost visibilityThe release adds Markdown export, session forking with archive and restore, asynchronous hooks that can invoke MCP tools, and estimated thread credit or cost display. It also adds Amazon Bedrock Runtime support and makes sandbox restrictions fail closed for denied or unreadable paths on Linux and Windows.→Cline Desktop 0.0.14 improves background agent executionThe release adds native macOS notifications, voice input, streaming command output, background execution for long-running commands, inline image generation, and collapsible summaries for completed runs. It also adds a separate Cline Code Beta application and fixes session-policy, skills, checkpoint, and command-execution issues.→WeWeb moves published-app CSS to browser-rendered stylesThe change is intended to improve editor and published-app performance, with the largest gains on larger projects and slower connections; WeWeb says internal tests roughly halved RAM use. The release note also positions the change as groundwork for making public pages easier for search engines and AI assistants to discover and understand.→Researchers show how Copilot can be induced to expose connected dataThe CoSnitch research found that repeatedly challenging Microsoft Copilot about an attack could reveal undocumented autorun and prompt-parameter behaviour. A crafted URL could then execute a malicious prompt in an authenticated session, potentially exposing connected application data or altering Copilot memory.→CISA orders urgent remediation for actively exploited Ray RCE flawCISA said attackers are exploiting CVE-2025-62593, a critical Ray vulnerability with a CVSS v4 score of 9.4. The browser-based attack can reach developer machines or private networks through DNS rebinding, and Ray 2.52.0 contains the fix.→Apple patches an image-processing flaw suitable for spyware deliveryCVE-2026-65346 is an integer-overflow flaw in ImageIO reported by Meta Red Team X and affecting Apple operating systems. Apple’s update also addressed multiple WebKit issues and a Telephony flaw that could allow network-traffic interception from a privileged network position.→ValidateHTML finds pervasive markup and accessibility failures across top domainsValidateHTML found that about 87.2% of the domains had invalid HTML, alongside 100,305 HTML violations and 18,863 CSS errors. More than one-third failed accessibility checks, including missing alternative text and ARIA labels.→Apple resets EU app economics for web and alternative-store distributionThe revised terms introduce a flat 5% commission on digital goods sold through apps distributed outside the App Store or through the web. Apple also set standard rates of 26% for in-app purchases and 20% for alternative payments, while loosening financial qualification rules for alternative marketplaces.→Warp packages the software-factory model for coding agentsWarp Factories packages triage, specification, implementation, review and verification into an agent workflow. It supports Codex and Claude Code, connects to Linear, Jira, Slack and Teams, and exposes performance and token-spend metrics; Warp’s CEO said 30–35% of the company’s weekly tasks are automated.→Rapid7 says AI is compressing the vulnerability-to-exploitation cycleRapid7 reported that high- and critical-severity vulnerability disclosures doubled year on year to 8,539 in Q2 2026, while newly exploited vulnerabilities rose 8% to 40. The analysis links the pressure to AI-assisted discovery and exploitation, as well as insecure code produced through vibe coding, and recommends prioritising exposure over severity scores alone.→Relay shutdown sends its founder back to Google Chrome with an agentic briefRelay’s shutdown update says free access ended on August 15 and paid access will end on September 14. TechCrunch reports that Bank is returning to Google and said Chrome has ambitious plans for working with AI and agents.→Bluesky attributes a day-long service disruption to a DDoS attackThe social network confirmed that a day-long disruption resulted from a distributed denial-of-service attack and said it was monitoring upgraded protections. Bluesky provided no technical details, while researchers cited by TechCrunch said Iran-backed attackers claimed responsibility without establishing a link.→Xpander raises $7.5 million to govern and operate AI agentsXpander’s platform includes a vendor-neutral agent harness that executes agents as portable workloads and renders their interfaces. Its governance layer covers products, workflows and data, while an agentic forward-deployed engineer named Omni is intended to accelerate adoption.→GitHub outage disrupted repository downloads and CopilotGitHub opened the incident at 13:40:03 UTC after failures affected GitHub.com services, with repository-content downloads later reported at roughly a 50% error rate. The status page continued to record Copilot authentication failures and residual impact during the window.→Linux 7.2 ships after an unusually large, AI-heavy development cycleLinux 7.2 was released after Linus Torvalds described the development cycle as unusually large and said AI tools were contributing to review and bug discovery. The kernel adds cache-aware scheduling and support work for Apple M3 and newer graphics and chip platforms, while deprecating AppleTalk and PCMCIA.→Claude API and Claude Code outage interrupted multiple Anthropic servicesAnthropic’s status API recorded the incident starting at 21:58:56 UTC on 16 August and resolved at 22:34:39 UTC. The affected components included the Claude API, Claude Code and the platform console, so both end-user and developer-facing surfaces were involved.→Wiz discloses an autonomous attack agent exploiting a Copilot-checked Snowflake PRWiz reported that its Red Agent used an issue-title injection to execute commands and obtain Snowflake Jira credentials from a public repository. The disclosure says Copilot Autofix co-authored the merged change and marked it clear, while Wiz clarified that it cannot establish whether the code change itself was AI-assisted.→Claude Opus 5 and Sonnet 5 suffered a separate developer-facing degradationAnthropic opened a separate incident at 13:56:41 UTC for elevated errors on Claude Opus 5, later reporting that Sonnet 5 was degraded as well. Claude.ai, the API, Claude Code and Claude Cowork were marked operational again at 15:29:25 UTC.→Cursor’s coding agents degraded when using non-fast Grok 4.6Cursor’s status API recorded the incident at 09:32:24 UTC, affecting Automations, CLI, Cloud Agents, cursor.com and the IDE. Cursor said users could see slowness and high error rates, while other models were unaffected, and marked the incident resolved at 13:40:26 UTC.→Vercel Dashboard disruption affected observability, analytics and firewall functionsVercel opened the incident at 14:30:28 UTC, affecting Dashboard observability, Web Analytics, Speed Insights, Firewall and Usage, with TOTP login issues also reported. By 20:13 UTC, services had recovered but Vercel was still ingesting missing observability data.→Agentic AI workflow costs forecast to rise more than fivefold by 2028The Register reported Gartner’s forecast that agentic AI workflow costs will increase by more than five times by the end of 2028. The explanation is that complex workflows consume more tokens and orchestration effort, so falling per-token prices may not offset total workload growth.→Cloudflare investigated network-performance issues in Phoenix and Los AngelesCloudflare opened the incident at 19:34:15 UTC, reporting network-performance problems in Phoenix and Los Angeles. The status page said the fix was being monitored at 20:34 UTC and resolution followed at 20:44 UTC, after the requested freshness window had closed.→Cloudflare reported elevated Durable Objects errors in SingaporeCloudflare recorded the incident at 07:47:04 UTC and said Durable Objects and downstream services in Singapore were returning increased errors. The provider resolved the incident at 08:58:06 UTC.→Cursor officially joins SpaceXCursor says the acquisition process begun in April is complete and that it will remain focused on making advanced models useful in software development. The company says SpaceX will provide access to substantial GPU capacity and that Cursor has begun testing Grok 4.6.→Cursor announces AIUC-1 agent-security certificationCursor says Schellman conducted an organisational-controls audit and adversarial testing across thousands of scenarios. The assessment covered secrets protection, secure code generation, MCP security, and agent identity and permissions, with recurring quarterly checks and an annual full audit planned.→Anthropic details Claude text watermarkingAnthropic says the method changes low-stakes token choices without adding hidden characters, extra tokens, or generation cost. It expects executable code to be less affected because exact output matters, while comments and natural-language material may show stronger watermarking.→ChainDrop targets npm users through developer-tool configurationThe Register reports that ChainDrop poisoned 444 npm packages and can activate through configuration used by environments such as VS Code and Claude Code. The campaign is designed to harvest npm tokens, cloud credentials, and other secrets, while dependency scanning may not detect the configuration-based payload.→DeepSeek open-sources a plugin-oriented agent harnessThe DeepSeek Harness is built on the Cordis framework and provides components for context handling, tool use, orchestration, state, permissions, and error management. It also supports append-only session logs that can be inspected, forked, searched, and replayed.→Claude Code users report blank reasoning blocks despite billingThe Register reports that Opus 4.8 and Sonnet 5 requests can return missing or incomplete thinking summaries. Anthropic documentation states that thinking tokens remain billable even when the thinking content is collapsed or redacted, and the company was investigating the behaviour.→Autonomous AI attacks raise critical-infrastructure concernsReporting describes concerns that increasingly autonomous AI systems could accelerate attacks against critical infrastructure and help turn digital compromise into physical disruption. The scale and likelihood of such attacks remain uncertain, but the risk depends directly on how software agents are connected to operational systems.→SQLite WAL-reset bug explains Tailscale outagesThe bug can occur when multiple connections write to and checkpoint the same SQLite database in WAL mode at nearly the same time. SQLite says the issue was present from version 3.7.0 through 3.51.2 and is fixed in newer releases.→Mojo reaches its 1.0 milestoneThe Register reports that Mojo has reached the 1.0 milestone as Modular continues operating under Qualcomm ownership. Developers are still waiting for the compiler to become fully open source, which affects how independently the language can be adopted and maintained.→celld brings Durable Objects-style state to self-hosted infrastructurecelld uses JavaScript-compatible Workers and Durable Objects interfaces while moving storage and coordination onto infrastructure controlled by the operator. The design combines object-level SQLite state with object storage such as S3-compatible systems.→Qwen3.8-27B expands locally deployable coding modelsQwen’s model card describes native 262,144-token context, extensibility to 1 million tokens, tool-use support and compatibility with Transformers, vLLM and SGLang. It reports scores of 61.7 on SWE-bench Pro, 73.0 on Terminal Bench 2.1 and 79.0 on QwenSWEBench; these are vendor-reported results awaiting independent validation.→DeepSeek releases a plugin-oriented agent harnessDeepSeek Harness treats models, tools, skills, sessions, sandboxes, filesystems, loops, orchestration and the user interface as plugins powered by the Cordis framework. The project is available through npm or source, but its README warns that compatibility-breaking changes are expected during the developer preview.→npm moves risky installation behaviour behind explicit approvalThe new defaults block preinstall, install and postinstall scripts, implicit node-gyp builds, Git dependencies and remote tarballs unless developers explicitly allow them. npm’s migration guidance uses project allowlists and also begins deprecating sensitive two-factor-authentication-bypass granular access tokens.→ChainDrop exposes persistence through Claude Code and VS Code configurationMicrosoft’s analysis describes more than 400 affected packages, a malicious preinstall payload, credential theft from developer and CI environments, and automated republishing through stolen npm access. The malware can also use GitHub credentials to inject Claude and Visual Studio Code configuration files into repositories, creating another path between compromised projects and developers.→Yadda 3.0 modernises JavaScript BDD for agent-assisted maintenanceYadda 3 removes browser bundling and obsolete integrations, adopts node:test, Biome and lefthook, adds Playwright and Puppeteer examples, and now ships TypeScript definitions. Its maintainer says Claude Code with Opus 4.8 performed most of the modernisation while a comprehensive test suite and separated implementation/test changes supplied the main constraints.→GraphQL tooling converges on LLM-generated mocks without a standardExpedia’s mockql-rs generates data for annotated fields at request time while forwarding real fields upstream, allowing partially implemented schemas to support development and testing. The GraphQL RFC remains a stage-zero strawman, but its current design requires mock validation in application test suites and includes guidance for coding-agent skills.→Deltix turns plain-English mobile tasks into replayable testsThe product runs locally against an iOS Simulator on a Mac while keeping application source, builds and signing identities on the developer’s machine. Successful exploratory runs become deterministic Playbooks that can be replayed against new builds, while Android support and a CI CLI are listed as forthcoming.→ACE Sidecar adds local cost and workflow observability for coding agentsThe sidecar records per-turn token, cache, latency and cost data, reads existing local transcripts, and stores metrics in a local SQLite database. It also exposes Prometheus metrics and mines repeated command sequences into installable SKILL.md files, while its optimisation recommendations currently measure rather than automatically change workflows.→Mole applies budgets and evidence checks to deep researchMole reserves every model call against a ledger before execution, checks extracted claims against verbatim source text, and records unsupported claims rather than silently retaining them. It runs as a static binary, supports OpenAI-compatible endpoints and restricts local-data analysis to aggregate outputs.→BriskDB experiments with sharded SQLite as a developer-facing databaseThe project uses independent SQLite WAL files behind a protocol-neutral Rust engine with shard-safe ID allocation, cross-shard indexes and operational endpoints such as health and metrics. Its README explicitly labels the software alpha rather than production-ready, with generated keys and cross-shard indexes still experimental or opt-in.→