AI Agent News from Official Blogs and Changelogs · page 5
-
Introducing the Microsoft Copilot harness in Foundry Agent Service
Microsoft introduces a Copilot harness in Foundry Agent Service, providing a unified execution foundation for agents with context, tools, and control management.
-
Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents
GitHub's 2026 Octoverse report shows record growth in commits and pushes as developers build with agents.
-
Route and orchestrate models with GitHub Copilot Fusion
GitHub Copilot Fusion is now generally available and intelligently routes and orchestrates models for different tasks.
-
GitHub Copilot app, Universe releases
GitHub Copilot app and Universe releases enable cross-repository work and access to specs, designs, and assets.
-
GitHub Enterprise now includes access to GitHub Copilot
GitHub Enterprise now includes GitHub Copilot access for all licensed users through GitHub or Microsoft accounts.
-
GitHub Copilot in VS Code, Universe releases
GitHub Copilot in VS Code now offers expanded model choice and control over where agents run.
-
GitHub Agentic Workflows generally available
GitHub Agentic Workflows is now generally available for automating reasoning-based tasks like issue triage and CI failure analysis.
-
OpenAI and ServiceNow expand partnership across enterprise workflows
OpenAI and ServiceNow expanded their partnership to integrate OpenAI models and agents into ServiceNow, and ServiceNow workflows into ChatGPT and Codex.
-
Legora is building the agentic operating system for legal work
Legora uses OpenAI models and ChatGPT to accelerate legal workflows and contract analysis.
-
Vercel Agent can now validate alerts before paging
Vercel Agent can now validate monitoring alerts and filter them before sending to paging systems, allowing only Critical-classified alerts to reach destinations like incident.io.
-
Run Paseo agents on Vercel Sandbox
Paseo now supports running multiple coding agents including Claude Code and OpenCode in Vercel Sandbox, allowing remote control across devices with persistent file and conversation history.
-
Vercel Plugin now available in OpenCode
Vercel Plugin is now available in OpenCode with specialist agents for deployment, performance optimization, and AI application design.
-
Agents can now enable Web Analytics and Observability Plus with Vercel CLI
Vercel CLI updated to allow agents to enable Web Analytics and Observability Plus without requiring interactive terminal input.
-
GitHub Copilot supports server-to-server authentication
GitHub Copilot SDK now supports server-to-server authentication for GitHub Apps, enabling services and agents to make requests on behalf of organizations.
-
Retell uses OpenAI models to help power 100M monthly calls
Retell voice agent platform uses OpenAI models to handle 100M monthly calls and has reached $1.6M annual revenue per employee.
-
Introducing n8n 3.0
n8n 3.0 beta released with focus on reliability, stability and security.
-
Vercel Agent sessions now cost 67% less
Vercel reduced Agent session costs by 67% through improved context management and better prompt-cache reuse, with updated CLI skills for workflows.
-
HarnessAgent now supports image inputs for all harness adapters
Vercel's AI SDK HarnessAgent layer now supports image inputs across all harness adapters, allowing users to pass images with text in standard message formats.
-
GitHub Copilot assisted permissions generally available
GitHub Copilot's assisted permissions feature is now generally available across Copilot CLI, app, and VS Code, allowing users to grant or deny permissions when Copilot requests to run commands or access file system paths.
-
Greptile takes bug detection beyond code review with GPT-6 Astra
Greptile uses GPT-6 Astra for testing user workflows, Daybreak for security, and Codex to reduce rework in bug detection.
-
GPT-6.1 Terra in GitHub Copilot
OpenAI's GPT-6.1 Terra model is now generally available in GitHub Copilot for coding and terminal work at lower cost than GPT-6.1 Sol.
-
Introducing GPT-6.1 Terra
OpenAI released GPT-6.1 Terra, a fast model for computer use, visual reasoning, and writing tasks.
-
Groundhog Day for AI agents: Prompt caching patterns for using OpenAI models on Amazon Bedrock
Amazon Bedrock now supports prompt caching for OpenAI models, reducing input costs and latency by caching agent context, tools, and instructions across steps.
-
Build agentic retrieval on a Managed Amazon Bedrock Knowledge Base with CDK
AWS CDK template demonstrates building an agent for retrieval across multiple Amazon Bedrock Knowledge Bases with routing, grounding, and observability.
-
GPT-6.1 Sol helps CodeRabbit strengthen security across complex codebases
CodeRabbit uses GPT-6.1 Sol to investigate security risks in codebases, with a security agent helping developers review findings and fixes.
-
Vercel Sandbox commands now support stdin
Vercel Sandbox now supports stdin for running commands, enabling communication with stdio MCP servers like agent-browser.
-
GPT-6.1 Terra now available on AI Gateway
GPT-6.1 Terra is now available through Vercel's AI Gateway with improvements in visual reasoning and coding for document and desktop application tasks.
-
Usage metrics API adds usage from tools like OpenCode
GitHub Copilot's usage metrics API now includes model usage data from additional tools like OpenCode that were not previously reported.
-
Turn notifications into code with GitHub Copilot in Teams
GitHub released a new Microsoft Teams app integrating GitHub notifications with GitHub Copilot to help teams convert repository updates into code actions.
-
Send feedback from the Vercel MCP
Vercel MCP now includes a submit_feedback tool for sending bug reports and feature requests to Vercel from agents.
-
How Vercel Agent accesses production safely
Vercel discusses how its agent safely accesses production systems by managing permissions and approval workflows for agent actions.
-
How Bayer Crop Science built an agentic decision engine on Amazon Bedrock AgentCore with AWS Unified Operations
Bayer Crop Science built a multi-agent platform on Amazon Bedrock AgentCore enabling natural language queries on analytics data for breeders and researchers.
-
AI Agent Memory: How LLM Agents Manage State in Production
Article on AI agent memory architectures and how to build workflows using n8n and Cognee that retain context across sessions.
-
LiteLLM: Claude Code can leak a GitHub CLI token into logged responses
LiteLLM's incident post on the 1.107.0 release says Claude Code can, in specific situations, put a user's GitHub CLI OAuth token into a response body, so gateways that log responses may store it. LiteLLM believes the attacker took such tokens from its own logs (as reported). No Anthropic statement found yet. Treat proxy logs of Claude Code traffic as secrets.
-
LiteLLM 1.107.0 carried a credential stealer; Railway and Zeabur report leaks of customer secrets
LiteLLM says its release pipeline was compromised and litellm 1.107.0 on PyPI and Docker carried credential-stealing malware; it advises pinning 1.106.2 and rotating keys, including provider keys stored in the proxy. Railway and Zeabur, as reported, had customer environment variables read by attackers seeking AI service keys. The post's wider list of 27 incidents was not checked by us.
-
GitHub Copilot in Visual Studio — September update
GitHub Copilot in Visual Studio added features including pull request exploration in chat, team model selection, and vulnerable package fixes.
-
Anima helps clinics cut call waits by 75% with GPT-6 and GPT-Live
Anima reduced clinic call wait times by 75% using OpenAI's GPT-6 and GPT-Live models to expand care team capacity.
-
Kimi K3.1 Pro and Kimi K3.1 Pro Highspeed are available now in AI Gateway
Kimi K3.1 Pro and K3.1 Pro Highspeed from Moonshot AI are now available on Vercel's AI Gateway, with improvements in coding, 3D modeling, financial analysis, and agentic tasks.
-
Turning Database Grading into RL Rewards: Post-Training Agents with ThinkingBox
ThinkingBox enables post-training of agents using database grading results converted into reinforcement learning rewards.
-
Custom skills, agents, and MCP servers in usage metrics API
GitHub Copilot usage metrics API now displays names of custom skills, agents, MCP servers, slash commands, and plugins instead of grouping them.