TAG
#mcp
Stories connected through a shared idea, actor, technology or theme.
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.→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.→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.→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.→Atlassian introduces AMP for coordinated human-agent collaborationAMP gives agents explicit identity, ownership, permissions and shared presence across Jira, Confluence, Loom and related collaboration surfaces. Atlassian also introduced an MCP server that exposes enterprise context to external AI tools and coding agents.→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.→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.→Predactiv details a behavioral AI foundation model for marketingThe model uses more than 65 billion monthly behavioural signals to create custom audiences, embeddings and plain-language insights. It is available through Predactiv’s platform, API and MCP server for use with major AI tools.→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.→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.→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.→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.→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.→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.→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.→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.→dv01 launches agentic infrastructure for structured financeThe platform includes agents for DealStudio and Credit Facility Management that can prepare analyses from offering documents, credit agreements and loan data. Its MCP server lets external agents call dv01's governed data and analytical capabilities instead of interpreting raw loan data independently.→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.→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.→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.→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.→Dun & Bradstreet exposes governed commercial intelligence through agentsThe service connects business identity, relationships and credit intelligence from the D&B Commercial Graph to Claude, ChatGPT, Codex, Copilot and Databricks integrations. The company says the system is designed to produce consistent, traceable and auditable outputs for finance workflows.→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.→Google Home opens device control to third-party AI agentsMCP-compatible agents including Claude, ChatGPT, Hermes, OpenClaw and Google Antigravity can access Google Home event history and control supported devices. Initial access is limited to Google Home Premium Advanced subscribers in the United States, with safeguards including a prohibition on agent-initiated door unlocking.→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.→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.→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.→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.→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.→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.→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.→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.→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.→Lightsage raises $4 million to measure how AI agents use softwareThe platform simulates agents using APIs, SDKs, command-line tools, MCP servers and agent skills, then identifies failures in documentation, authentication or execution. Nexus Venture Partners led the round, which Lightsage said will fund broader evaluation, analytics and attribution capabilities.→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.→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.→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.→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.→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.→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.→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.→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.→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.→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.→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.→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.→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.→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.→Anthropic previews a hardware standard for AI agentsAnthropic released the Model Hardware Standard, a model-agnostic framework for connecting AI agents to equipment including microscopes, manufacturing machines, quantum hardware and robot arms. The framework builds on MCP and is being shared with partners while Anthropic develops safety evaluations ahead of a planned open-source release.→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.→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.→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.→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.→Google Cloud launches Gemini Enterprise solutions for legal and financial servicesGemini Enterprise for Legal and Gemini Enterprise for Financial Services entered preview with purpose-built agents, MCP connectors, domain skills and centralised audit and permission controls. The financial-services release includes more than 50 skills and 13 connectors, while the legal offering targets contract review, diligence, regulatory monitoring and privacy workflows.→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.→Britive launches task-scoped runtime controls for AI-agent accessARC is designed to grant intent-bound access during a job and remove that access afterwards. Britive says its controls cover identity, requests, authorisation, privileges, actions and outcomes, but the release is a vendor product claim rather than independent evidence of deployment or effectiveness.→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.→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.→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.→TransformativeMed and Red Rover put agentic discharge planning inside Oracle HealthTransformativeMed said its Agentic Clinical Operating System will combine with Red Rover Health's interoperability layer and Model Context Protocol connections to access clinical context and workflow tools. Their first joint project is agentic patient-discharge planning with an academic health system, with a demonstration planned for the Oracle Health Summit in September.→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.→Binance opens trading and account actions to AI agents through Agent OSThe platform supports compatible agents through MCP, with user-configured permissions, dedicated subaccounts and revocable access; the initial implementation can read market and account data and place trades. Binance says it monitors resulting orders, but the selected AI application's broader workflow and reasoning remain outside Binance's visibility.→Thomson Reuters makes CoCounsel Legal generally available as an agentic workflowBuilt on Anthropic's Claude Agent SDK, CoCounsel Legal can move from legal research and issue analysis to a first-draft brief while checking cited authority. Thomson Reuters also announced an expanded MCP connection with Claude and said one million professionals in 107 countries and territories rely on CoCounsel.→Redpanda adds an out-of-band policy engine at the MCP boundaryIn its second-quarter fiscal 2027 results, Redpanda announced an out-of-band policy engine for the MCP boundary and a live graph of agents, models and tools. It also reported new annual recurring revenue growth of more than 100% year on year and general availability of its SQL engine on Google Cloud and AWS; those business metrics are company-reported.→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.→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.→Aiva says its nursing assistant can control more than 40 hospital IT systemsAiva said its nursing assistant now uses ChatGPT for Healthcare, Claude for Healthcare and Gemini Enterprise behind one voice interface for more than 40 hospital IT systems. The company cited integrations with Epic, Oracle Health and ServiceNow, alongside clinical reference, communication, nutrition and smart-room functions; the release provides no independent clinical evaluation.→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.→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.→