Runtime
Sessions, event logs, crash recovery, harness interfaces, sandbox adapters, and API routes are in the repo.
Open-source Claude Tag
Open Managed Agents is not a finished clone of Anthropic's product. It is the Apache 2.0 runtime layer you would use to build one: Slack publication, durable sessions, MCP tools, vault-backed credentials, memory stores, sandboxed execution, model cards, and recovery.
Sessions, event logs, crash recovery, harness interfaces, sandbox adapters, and API routes are in the repo.
Slack, GitHub, Linear, MCP, vaults, model cards, memory stores, and custom integration patterns are open and extensible.
Run it in your Cloudflare account, with Docker, or on the Node.js backend instead of depending on a closed hosted runtime.
The honest positioning matters for SEO and trust. Open Managed Agents provides the foundation, but a full Claude Tag-level product still needs a dedicated setup console, access-bundle UI, channel memory controls, spend limits, audit dashboards, and ambient-mode governance.
See the feature comparisonAnthropic's Claude Tag is not open source. Open Managed Agents is an Apache 2.0 project that provides an open-source runtime for building Claude Tag-style agents.
The runtime can cover the agent publication, durable sessions, tool access, MCP integration, vault-backed credentials, memory primitives, sandboxed work, and recovery. Product polish such as the Tag admin console is still roadmap work.
Yes. Open Managed Agents can run on Cloudflare Workers, Docker, or Node.js, with your model keys and your integration credentials.