Anthropic publishes security hardening guide for Claude Managed Agents deployments
brada · x · 2026-10-14
Anthropic engineer brada announced a published guide for locking down Claude Managed Agents deployments, emphasizing that secure cloud-hosted agents aren't a weekend project — API keys live in a vault and the sandbox only ever sees a placeholder.
Key settings covered, in priority order:
- Default to cloud sandboxes unless data must stay on your infrastructure
- networking.type: limited with only required hosts
- Per-credential injection location and host restrictions
- alloweddomains allowlists on webfetch/websearch
- permissionpolicy: auto for read-only tools, alwaysask for irreversible actions
- Smallest toolset: default disabled, enable as needed
- No agent reading untrusted input holds an always-allow write/execute tool
- Session and deployment budget caps; optional offline transcript reviewer
Related event: Anthropic Releases Security Hardening Guide for Claude Managed Agents(2 posts)→
More from coding & agent
- Google offers 10,000 scientists free 6-month top-tier Antigravity plans — danielrock · 2026-10-14
- CineRecon Benchmark: 12 Frontier Models Struggle to Reconstruct Editable 3D Scenes from Video — hkuhk · 2026-10-14
- Meituan's ClawLoop: Self-Improving Agents via Teacher-Free Harness-Native Training Loops — meituan-longcat · 2026-10-14
- Grok Bot's big leap: full breakdown of latest updates and autonomous business test — elonmusk · 2026-10-14
- Indie dev cuts homepage job list load from 1.8s to 0.1s in performance push — DanielLockyer · 2026-10-14
- Treat Your AI Harness Like Code: a CI Step That Ships Only Winning Candidates — AKAMA199 · 2026-10-14