Docker
7 min
the official image is published to github container registry ghcr io/litmusautomation/litmus mcp server\ latest sse only (mcp server, no web ui) the simplest deployment runs only the mcp server on port 8000 docker run d name litmus mcp server \\ p 8000 8000 \\ ghcr io/litmusautomation/litmus mcp server\ latest the sse endpoint is at http //localhost 8000/sse and a health check is at http //localhost 8000/health with the web ui expose port 9000 and provide at least one llm api key docker run d name litmus mcp server \\ p 8000 8000 p 9000 9000 \\ e anthropic api key=\<key> \\ ghcr io/litmusautomation/litmus mcp server\ latest open http //localhost 9000 in your browser see web ui # for what's available there you can also enter the api key later from the web ui's setup screen start without e if you prefer arm64 (apple silicon and others) the image is built for linux/amd64 on arm64 hosts, pin the platform docker run d name litmus mcp server \\ \ platform linux/amd64 \\ p 8000 8000 \\ ghcr io/litmusautomation/litmus mcp server\ latest splitting web ui and mcp server across hosts if you deploy the web ui on one host and the mcp server on another, point the web ui at the server e mcp sse url=http //\<mcp server host> 8000/sse environment variables variable default purpose anthropic api key unset anthropic claude api key (web ui) openai api key unset openai api key (web ui) gemini api key unset gemini api key (web ui) mcp sse url http //localhost 8000/sse where the web ui looks for the mcp server enable stdio false set true for stdio mode see stdio mode # llm keys are persisted to /app/ env once entered through the web ui, so you typically only need to pass them once docker compose services litmus mcp server image ghcr io/litmusautomation/litmus mcp server\ latest ports \ "8000 8000" \ "9000 9000" environment anthropic api key ${anthropic api key} volumes \ /opt/litmus mcp/ env /app/ env the volume mount is recommended; see persistent configuration # next persistent configuration # keep settings across container replacement clients # wire an mcp client into the running server headers reference # what to send with each request