Daily · · Claude · 🌶️🌶️Worth a lookMedium: worth a look Why: Agent history stays on your servers
Self-hosted Claude Managed Agents, your server
Your agents’ session history can now live on your own server.
Claude Platform’s Nov 12, 2026 release notes launch self-hosted Claude Managed Agents, where you run the whole service and not only the sandboxes. Access goes through the Anthropic account team and isn’t self-serve.
What changed
You host the service, the sandboxes and the database, while the model can still run at Anthropic or at a provider you choose. Anthropic’s docs say the server stores agents, sessions and each session’s history, and sends model requests to Anthropic by default.
Both the hosted and self-hosted versions share the same API, so the SDK and the ant CLI work with your server.
What you run
- A server (
cma-server) that serves the API and runs the agent loop, plus a PostgreSQL 16 or later database. Without a database, a restart erases the data. - Workers that run the agent’s shell and file tools in your sandboxes. The docs warn that the bash tool is a real shell, so isolation comes from the container or VM the worker runs in.
- A Console web page on your own server, where people sign in through your identity provider.
Who it is for
- Teams that must store agents, sessions and history themselves.
- Teams whose model calls must go to Amazon Bedrock, Claude Models on Google Cloud (Google managed), Microsoft Foundry or their own model gateway.
- If you only need tools to run on your infrastructure, Anthropic points you to self-hosted sandboxes instead.
What to do
- Decide whether you need the whole service on your side or only the sandboxes.
- To request access to the download, contact your Anthropic account team. It is not self-serve.
- Read “Differences from the hosted service” first, because some features exist only on the hosted service. You also install each release yourself.
Source: Claude docs: Self-hosted Claude Managed Agents · Claude Platform release notes, Nov 12, 2026