AI Agent News from Official Blogs and Changelogs · page 4
-
Claude Code: Haiku 5.5 now writes background subagents' status lines, about 5% less limit used
Anthropic says a background subagent's status line, refreshed every 30 seconds, used to re-read the whole conversation; Haiku 5.5 now writes it from short notes, about 99% cheaper per update (as reported). The subagent keeps its own model. Sessions with background subagents used about 5% less of the limit, roughly 6% more prompts.
-
Give Your Muse a Voice
We’re rolling out voice and expressive avatars in the Muse app, giving you a new way to interact with your Muse. The post Give Your Muse a Voice appeared first on Meta Newsroom .
-
Agents aren’t humans (but they kinda are)
Cloudflare discusses how AI agents appear to servers as intermediaries that obscure user visibility, and proposes collaborative methods for identifying and managing automated agent traffic safely.
-
Bring Agentic Coding On-Device with Foundry Local
Microsoft Foundry Local enables agentic coding to run on developer machines locally.
-
GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents
OpenAI introduced GPT-6 Cyber, a model designed to find exploitable vulnerabilities and develop fixes while staying within authorized scope.
-
Merge builds product and business tools with OpenAI technology
Merge uses OpenAI models and Codex for product code review, marketing support, and business planning tasks.
-
Build Live Apps in Cursor
Cursor now supports building and deploying web apps with cloud agents handling hosting, database, auth, and deployment.
-
GitHub Copilot is now available in Xcode’s coding assistant
GitHub Copilot is now available as a coding assistant in Xcode, currently in public preview.
-
Kimi K3.1 Pro now has more ZDR providers on AI Gateway
Kimi K3.1 Pro models are now available through multiple Zero Data Retention providers on Vercel's AI Gateway, expanding beyond Moonshot AI.
-
Four layers of context engineering for production AI agents: How Deep Insight applies them in working code
AWS post describes a four-layer context engineering framework for production AI agents, implemented in Deep Insight, an open source multi-agent data analysis solution using Strands Agents SDK and Amazon Bedrock.
-
How Indegene is modernizing and scaling enterprise data quality in life sciences
Indegene built an LLM-powered Data Quality Agent on AWS that generates context-aware data quality rules and produces SQL and Python code under human-in-the-loop governance for life sciences organizations.
-
How U Mobile built agentic telco operations with Amazon Bedrock AgentCore
U Mobile deployed AiDaH, an agentic operations platform on Amazon Bedrock AgentCore, handling over 150 operational scenarios and reducing trouble tickets by up to 84.5% nine months after launch.
-
See the agents, set the terms: New visibility and control for personal agents
Cloudflare announced tools for site owners to see and control personal AI agent access to their websites.
-
DevRev delivers autonomous self-service with GPT-6.1 Sol
DevRev uses GPT-6.1 Sol to power Computer and Operator agents for enterprise self-service, balancing reasoning, speed, and cost.
-
Dependabot version updates support Agent Package Manager
Dependabot now supports version updates for Agent Package Manager, allowing automated updates to AI agent skills, prompts, and instructions.
-
Agentic cheminformatics with Google Antigravity
Google Antigravity accelerates drug discovery workflows with cheminformatics skills for searching compounds and proposing novel macrocycles.
-
Build a semantic layer for agentic AI with EDB Postgres AI and Amazon Bedrock AgentCore
EDB Postgres AI and Amazon Bedrock AgentCore enable building a semantic layer for agentic AI with governed access control for Strands Agents.
-
Same Model, Different Harness: CUGA, Hermes, and Deep Agents — When to Pick CUGA?
Comparison of agent frameworks CUGA, Hermes, and Deep Agents discussing when to use each.
-
GitHub Copilot is now available in JetBrains Air
GitHub Copilot is now available in JetBrains Air, using the Agent Client Protocol for coding task delegation and execution.
-
Vercel MCP now supports Client ID Metadata Documents
Vercel MCP now allows any client to connect by publishing a Client ID Metadata Document, removing the need for manual review and allowlisting.
-
Prepare web content for Amazon Bedrock Knowledge Bases with Crawl4AI
Guide on using Crawl4AI web crawler to prepare web content for Amazon Bedrock Knowledge Bases, handling JavaScript rendering and metadata enrichment.
-
Vercel MCP now renders deployments as an interactive card
Vercel MCP now displays deployments as interactive cards in ChatGPT and other MCP-compatible chat clients, showing status, logs, and deployment details.
-
Build efficient agents with knowledge distillation on Amazon SageMaker AI
AWS SageMaker AI enables knowledge distillation to transfer agentic behavior from expensive frontier models to smaller, lower-cost student models for production-scale agent deployment.
-
Agentic engineering intelligence with Amazon Nova Multimodal Embeddings and Amazon Bedrock AgentCore
Multi-agent engineering assistant using Amazon Nova Multimodal Embeddings, Bedrock AgentCore, and Strands Agents SDK reduces aerospace component substitution evaluation from days to minutes.
-
Introducing Vercel Security Review
Vercel launched Security Review in public beta, a new capability in Vercel Agent that scans for vulnerabilities on GitHub repository pushes and reports findings as pull request comments.
-
Qodo builds context into AI code review with Codex
Qodo integrated with Codex to add company knowledge and coding standards to AI code review, helping developers catch critical issues before pull requests.
-
ai-design-check: 76 design rules for UIs built with coding agents, now on SwiftUI and Compose
ai-design-check (MIT) checks interfaces built with Claude Code, Cursor or Codex against 76 design rules with no LLM or API key, and adds a skill that fixes the screen and re-checks it. It now reads SwiftUI and Jetpack Compose; its makers say their own iOS app took 16 fixes (as reported). The repo is days old.
-
Claude Code's prompt-audit proposes edits; it does not delete your prompts
A viral post says /claude-api prompt-audit deletes whatever slows Opus 5.5. Anthropic's skill file says it reports dated patterns and proposes a diff, never applying edits without consent, and is not about making prompts shorter. It arrived in Claude Code 2.1.221; for CLAUDE.md, skills and agents, 2.1.283 added /doctor prompt-audit.
-
Preview web apps live as your coding agent builds them with Amazon Bedrock AgentCore runtime
Amazon Bedrock AgentCore Session endpoints now support authenticated URLs that forward HTTP requests and WebSockets to agent runtime ports, enabling live preview of web apps as coding agents build them.
-
Vercel MCP now supports MCP events
Vercel MCP now publishes events, so agents can subscribe to Vercel platform events like deployment.succeeded and deployment.error and react to what happens on your projects instead of polling for…
-
Building intent-driven UI testing with Amazon Nova Act
Epsilon built a UI testing service using Amazon Nova Act and Amazon Bedrock, allowing tests written in plain English to run on AWS-managed browsers.
-
Build a repeatable PHP-to-Java migration engine with Kiro
Kiro enables repeatable PHP-to-Java migrations through an automated discovery scanner and eight-phase workflow that verifies parity with the original application.
-
How MoEngage turned over 400 APIs into agent tools with AgentCore Gateway
MoEngage converted 400 private REST APIs into permission-scoped agent tools using Amazon Bedrock AgentCore Gateway, avoiding per-service MCP servers and context bloat.
-
Accelerate agentic AI with NVIDIA Nemotron models on Amazon EKS and SageMaker HyperPod EKS
NVIDIA Nemotron 3.5 Super VL, a 120B-parameter mixture-of-experts model, can be deployed on Amazon EKS and SageMaker HyperPod EKS for agentic AI applications.
-
Why We're Moving n8n's Automation Team Beyond the Service Desk Model
Romina da Costa, head of internal AI and automation at n8n, shares the four principles her team uses to help every team build and own its automations.
-
Mistral Large 4 updated weights and lower pricing on AI Gateway
Mistral Large 4 received updated weights with improved agentic coding performance and lower pricing on Vercel's AI Gateway.
-
Welcome Le Chonk: Mistral Large 4 on Hugging Face
Mistral Large 4 model is now available on Hugging Face.
-
Vercel Plugin is now available in Antigravity
Vercel Plugin is now available in Google Antigravity marketplace, providing agent guidance on Vercel APIs and products including AI SDK, Functions, and Workflow.
-
Building an Agentic AI Platform on Amazon Bedrock AgentCore
AWS publishes architecture guide and open-source blueprint for building multi-account agentic AI platforms on Amazon Bedrock AgentCore, including control plane and CI/CD evaluation pipeline.
-
Accelerating software delivery with agentic QA automation using Amazon Nova Act – Part 3
Part 3 of QA automation series using Amazon Nova Act adds test organization tools and team-based access controls for managing agentic QA across organizations.