feature

Managed Agents permission policies support auto evaluation mode

docs.claude.comagent

Claude Managed Agents now support 'auto' mode in permission policies, allowing servers to automatically evaluate each agent or MCP tool call—running, denying, or pausing for approval. The ant CLI 1.32.0 adds ant beta:sessions connect for terminal-based session monitoring and approval workflows.

Stacks this affects

Claude ships in 34 tested stacks here.

Full-stack with Lovable
Code · Builder
$70/mo
Cloud IDE with Replit
Code · Builder
$40/mo
UI-first with v0
Code · Builder
$60/mo
AI newsletter
Writing · Newsletter
$80/mo
AI newsletter (Substack)
Writing · Newsletter
$53/mo
AI thumbnails + ad creative
Marketing · Visual
$50/mo
Android app with AI
Code · Mobile
$42/mo
Autonomous research agent
Research · Intelligence
$59/mo
Cheap bulk LLM automation
Code · Backend
$140/mo
Cold outbound
Sales · Outbound
$342/mo
Cold outbound (solo)
Sales · Outbound
$75/mo
iOS + Android with Expo
Code · Mobile
$67/mo
Faceless YouTube
Video · Creator
$92/mo
Faceless YouTube (budget)
Video · Creator
$6/mo
Inbox triage
Productivity · Inbox
$51/mo
Inbox triage (free)
Productivity · Inbox
$26/mo
Indie SaaS
Code · Builder
$75/mo
Indie SaaS (Gemini long-context)
Code · Builder
$75/mo
iOS app with AI
Code · Mobile
$48/mo
LinkedIn ghostwriting
Writing · Agency
$129/mo
LinkedIn (for yourself)
Writing · Personal
$33/mo
Live trend research
Research · Intelligence
$70/mo
Multilingual content factory
Writing · Content
$60/mo
Open-weight LLM stack
Code · Backend
$170/mo
SEO content engine
Writing · SEO
$118/mo
Antigravity (Google)
Code · Workflow
$40/mo
Codex (CLI + cloud)
Code · Workflow
$40/mo
Ship code with AI
Code · Workflow
$50/mo
Cross-platform short-form
Marketing · Distribution
$49/mo
Self-hosted scheduler
Marketing · Distribution
$25/mo
TikTok character
Video · Creator
$92/mo
TikTok character (audio + text)
Video · Creator
$42/mo
UGC ad creative
Marketing · Ads
$67/mo
UGC ads (CapCut-only)
Marketing · Ads
$24/mo
All Claude changes and pricing

More Claude

featureCache diagnostics graduated from beta on Claude APIfeatureCache diagnostics exits beta on the Claude APIreleaseClaude Opus 5.5 launched with 1M context and always-on adaptive thinkingreleaseClaude Opus 5.5 released for long-running agentic workfeatureTools can be added or modified mid-conversation without cache invalidationreleaseClaude Opus 5.5 launches with 1M token context and always-on thinking