Configuration

the litmus mcp server is configured at two layers per connection credentials passed as http headers (sse) or environment variables (stdio) the server is stateless; every tool call reads these on demand server side persistent state stored in /app/ env and managed via the web ui things like registered litmus edge instances, the active one, the preferred llm provider and model pages page layer what it covers headers reference # per connection the full list of headers the server recognizes, grouped by backend (litmus edge, nats, influxdb, lem) multi instance edge # persistent registering and switching between multiple litmus edge devices from one mcp server llm providers # persistent api keys and model selection for the web ui's chat where each setting lives setting source notes edge url , edge api client id , edge api client secret per request headers mirror of the active instance (see multi instance edge) nats per request headers token credentials from litmus edge system > access control influx per request headers datahub credentials firewall must permit port 8086 edge manager url , edge api token , etc per request headers lem tenant credentials edge instance {i} , active edge instance env managed by web ui anthropic api key , openai api key , gemini api key env (or container env) read by the web ui; not used by the mcp server itself preferred model , preferred model id env web ui model selection persistent settings survive container replacement only with a volume mount by default env is inside the container see persistent configuration # for the host side mount that keeps it