Claude Managed Agents get per-session overrides and lifecycle webhooks
Anthropic shipped two changes to Managed Agents. The new agent_with_overrides type lets you swap the model, system prompt, tools, MCP servers, or skills for a single run, so you can test one variation without redeploying the whole version.
Webhooks now cover the full lifecycle: agent, deployment, and deployment run. You can subscribe to a published version, a paused deployment, or a failed scheduled run instead of polling for it.
The Agent SDK also gets a fix for a bug that made MCP servers registered through the initialize control request wait for the next turn before connecting. The names “Claude Browser” and “Claude Preview” are now reserved, so user servers can’t take them.
Up next
Anthropic is lobbying for tougher AI laws, state by state
Anthropic was the only major AI lab to back California’s 2025 safety law, and now, per POLITICO, it’s pushing states toward ever stricter rules bill by bill instead of the single national framework OpenAI favors. Illinois just mandated annual independent safety audits.
Related stories
- Claude Managed Agents get per-agent effort levels
- Claude Managed Agents can now hand tasks off to each other
- Managed Agents get session budgets and an advisor model
- Anthropic hands Agent Skills to the foundation behind MCP
- Anthropic’s daily-brief agent treats a failed read as news
- Claude’s SDKs run the click loop, but the browser is yours
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.