Anthropic launches private beta of self-hosted Claude Managed Agents for on-prem deployment
AI寒武纪 · wechat · 2026-11-13
Anthropic has opened a private beta of Self-Hosted Claude Managed Agents, letting enterprises run the full agent stack — orchestration server, session database, and execution sandboxes — inside their own cloud or data center.
Key points:
- The orchestration server only handles requests and task scheduling, never executing commands directly; worker programs run shell commands inside isolated sandboxes; a mandatory enterprise PostgreSQL stores all configs and session history.
- Model calls are fully flexible: Anthropic API, AWS Bedrock, Google Cloud, Microsoft Foundry, or an in-house gateway — only the inference request leaves the network.
- Three-tier key design (local API key, model credentials kept only on the orchestrator, per-environment sandbox keys) ensures a compromised sandbox can't leak model credentials.
Anthropic warns workers are real terminals with real permissions: run them in ephemeral containers/VMs, destroyed after use. Supported on Linux and Apple-silicon macOS; access via the sales team.
Related event: Anthropic Launches Self-Hosted Claude Managed Agents in Private Beta(2 posts)→
More from coding & agent
- Intern shipped a multiplayer app built with Claude Code, got hired, then presented it on stage — Alec_Coughlin · 2026-11-14
- Dev: Working at a High Level with AI Is Great but Exhausting — jonathan_wilke · 2026-11-14
- Open-source bastyn statically scans MCP and coding-agent configs for risks — claudio-rootedlogic · 2026-11-14
- 'We'll forget how to use apps': devs say agents make hand-coding feel unbearable — robleclerc · 2026-11-14
- OpenComputer launches as open-source, self-hostable Devin alternative claiming 3.3x faster coding agents — zeeg · 2026-11-14
- 15 building blocks of AI agent memory, distilled from building 400+ agents — MaryamMiradi · 2026-11-14