Anthropic opens self-hosted Claude Managed Agents to customers in private beta
For the chief technology risk officer at a bank
Anthropic opened its self-hosted Claude Managed Agents to customers in private beta, according to an announcement the company published on its Claude resources site on Nov. 12, as reported by AI Buzz Wire.1
The sources describe the aim as letting enterprises keep the control plane for agents that run unattended on internal systems inside their own environment, rather than in a vendor's data center.1 Managed Agents first launched as a hosted service on the Claude Platform, and Anthropic later added self-hosted sandboxes; the new mode has customers also run the server, its database and the workers, AI Buzz Wire reported.1
Model requests can go directly to Anthropic, through Amazon Bedrock, Google's Agent Platform or Microsoft Foundry, or through the customer's own gateway, according to AI Buzz Wire's account of the announcement.1
Session history is stored in the customer's own database, and organizations with zero-data-retention agreements with Anthropic get the same treatment for requests to Claude as for any other Messages API call, AI Buzz Wire reported.1
Credentials stay outside the sandbox, with a proxy adding them to requests for MCP tools and supported API calls; customers can also keep credentials in their own secrets manager.1
According to AI Buzz Wire's account of the announcement, each environment can be restricted to an allowlist of hosts, and each tool call and its result is logged, with records that can be exported over OpenTelemetry.1
A bank can keep the agent loop, secrets and audit trail in its own perimeter, but prompts still leave for whichever model endpoint it picks, so data-handling terms with Anthropic or the cloud provider still decide what is retained. The bank would also own the logs and the work of reviewing them.
Access runs through Anthropic account teams, and the general availability timeline is unclear, according to AI Buzz Wire.1
Our take. Self-hosting settles where the agent's secrets and session records live: inside your walls. It also settles who runs the server, the database, the workers and the log review. The vendor-perimeter objection is gone. A staffing question replaces it.
Action. Technology risk: Add self-hosted agent server, vault, sandbox allowlists and OpenTelemetry session logs to the control inventory and security monitoring review before any pilot.
What our records say
Anthropic sells banks Claude seats and API access directly and through cloud marketplaces, and in 2026 added embedded engineering with Goldman Sachs and FIS, and its bank record spans Goldman Sachs, JPMorganChase, BankUnited and Flagstar.1 What banks do with Claude: Citizens' internal gen-AI portal lets staff choose Claude alongside OpenAI and AWS Bedrock models.2 U.S. Bank gives engineers Cognition and Claude Code after starting with GitHub Copilot.3
From our records
- Claude (Anthropic)vendor · banks using it (24) Goldman, JPMorgan, BankUnited · alternatives Microsoft Copilot, ChatGPT Enterprise, Gemini Enterprise
- Amazon Bedrock (AWS)vendor · banks using it (11) BankUnited, FNBO, First Interstate · alternatives Microsoft Copilot, ChatGPT Enterprise, Claude
- Microsoft Foundry (formerly Azure AI Foundry)vendor · banks using it (3) Northern Trust, BNY, Ally · alternatives Amazon Bedrock, ChatGPT Enterprise, Claude