Atomic Agent
Delegates a full agent turn to managed Atomic Agent through Chat Completions. This is a remote agent, not an MCP server.
What it provides
Hosted on Falkenstein with dedicated bot data and usage limits. No endpoint or API key needed. Model calls use your OpenClaw Launch credits. Enabled remote tools may change files or run commands. Disabling keeps your data for later re-enabling.
Connection requirements
Run a pinned, tested Atomic Agent release outside the bot, configure its backend/model there, and start `atomic-agent serve --host 127.0.0.1 --port 8787 --cwd /path/to/work --api-key "$ATOMIC_AGENT_API_KEY"` behind an HTTPS reverse proxy. Enter the HTTPS origin without /v1, keep Bearer selected, and paste that server API key. This preview HTTP service exposes OpenAI-compatible Chat Completions; it is not an MCP server. Each request runs a full agent turn with its enabled tools on the external machine. Choose a dedicated working directory and restrict enabled tools there; calls can change files or run commands under that agent’s permissions.
- Protocol
- OpenAI-compatible REST
- Endpoint
https://atomic-agent.example.com- Authentication
- bearer
get /healthpost /v1/chat/completions '{"messages":[{"role":"user","content":"Summarize the README in your configured workspace."}],"stream":false}'Where to add it
Open Dashboard Tools, select a running OpenClaw or Hermes instance, then use the card’s Install or Connect action. External services may require an HTTPS endpoint and credentials; the dashboard shows those fields before anything is saved.