Litmus MCP Server
Official Litmus Automation MCP server: 62 tools to configure and monitor Litmus Edge industrial gateways.
- 13 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The tools create and delete devices and tags, run Docker containers and control packet capture on an industrial gateway.
- The OAuth2 client secret and NATS and InfluxDB tokens give access to equipment data and the working gateway configuration.
Install
- Works in
Manual install
docker run -d --name litmus-mcp-server -p 8000:8000 ghcr.io/litmusautomation/litmus-mcp-server:latestREADME command. On ARM64 add `--platform linux/amd64`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The tools create and delete devices and tags, run Docker containers and control packet capture on an industrial gateway.
- The OAuth2 client secret and NATS and InfluxDB tokens give access to equipment data and the working gateway configuration.
This is third-party code. Review the repository files before installing.
What it does
The server connects an agent to Litmus Edge, an industrial platform for collecting equipment data. It carries 62 tools in 13 categories: DeviceHub devices and tags, NATS topics, historical InfluxDB data, system events, the firewall, packet capture, digital twins and Docker containers. With a Litmus Edge Manager address set, `lem_*` tools add licenses, groups and projects, and any call can be routed to a specific gateway through LEM. Three SDK tools, `litmus_sdk_discover`, `litmus_sdk_read` and `litmus_sdk_write`, open a catalog of about 550 functions, and writes need explicit approval. The Docker image also ships a web chat on port 9000.
Who it is for. Automation engineers at plants and integrators who set up Litmus Edge: Modbus and OPC UA drivers, tags, digital twins.
Good fit when
- A dozen controllers are connected to a gateway and tags are created in bulk from a description.
- Tags in Failed status across all devices have to be found to see what dropped.
- A week of machine data is pulled from InfluxDB to compute the mean and deviation.
- Gateways whose license expires within the month are looked up in a Litmus Edge Manager project.
Not a fit when
- The gateway host is ARM64 with no way to run an AMD64 image.
- A live plant with no test gateway: writes and containers change the working environment.
- The data sits in another SCADA system: the server talks only to Litmus Edge.
Example request
Show tags in Failed status across all DeviceHub devices and the mean temperature over the last day.MCP
- Transport
- http, sse
- Authentication
- API key
Environment variables
EDGE_URLrequired- Litmus Edge base URL including `https://`.
EDGE_API_CLIENT_IDrequired- OAuth2 client ID from Litmus Edge.
EDGE_API_CLIENT_SECRETrequired, secret- OAuth2 client secret from Litmus Edge.
NATS_PASSWORDsecret- Access token from System, Access Control, Tokens. Needed to read NATS topics.
INFLUX_USERNAME- DataHub user for InfluxDB queries.
INFLUX_PASSWORDsecret- DataHub user password.
EDGE_MANAGER_URL- Litmus Edge Manager URL for the `lem_*` tools.
EDGE_API_TOKENsecret- Litmus Edge Manager API token.
Limitations
The server is built for linux/AMD64, and on ARM64 the image runs with the `--platform linux/amd64` flag. NATS topics need a token from System, Access Control, Tokens, and history needs port 8086 opened in the gateway firewall. The web chat on port 9000 has no login and stores LLM keys and Litmus Edge data, so publish it only on a trusted network. Claude.ai and Claude Desktop reject self-signed certificates, so HTTPS needs a trusted authority.
How to disable. Stop the container with `docker stop litmus-mcp-server` and remove the server entry from `~/.claude/mcp.json`. Revoke the OAuth2 client secret in Litmus Edge.
FAQ
How do I run Litmus MCP Server?
Run `docker run -d --name litmus-mcp-server -p 8000:8000 ghcr.io/litmusautomation/litmus-mcp-server:latest`. Port 8000 serves `/mcp` for Streamable HTTP and `/sse` for older clients.
How do I connect it to Claude Code?
Add an entry of type `http` with the address `http://localhost:8000/mcp` and the headers `EDGE_URL`, `EDGE_API_CLIENT_ID`, `EDGE_API_CLIENT_SECRET` to `~/.claude/mcp.json`. NATS and InfluxDB headers are added when topics and history are needed.
What is needed to read history from InfluxDB?
Port 8086 open over TCP in the Litmus Edge firewall and DataHub user credentials in `INFLUX_USERNAME` and `INFLUX_PASSWORD`. Without them the history tools return a connection error.
Can the agent change the gateway configuration?
Yes: create devices and tags, delete tags, run Docker containers. Through `litmus_sdk_write` a write passes only with the `user_approved` argument, which the assistant sets after you approve.
Does the server have its own interface?
Yes, a chat on port 9000. It accepts an Anthropic, OpenAI or Gemini key and switches between several Litmus Edge gateways. It has no login, so the port is opened only on a trusted network.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server for driving Windows: the agent opens apps, clicks and types, reads window and UI state, and works with files and PowerShell
Official Mailtrap MCP server: send email, test in the sandbox, read delivery logs, manage templates, contacts and domains
The official Postmark MCP server: 24 tools to send email, manage templates, search messages and diagnose delivery.
Microsoft Power Platform MCP server with 312 tools for flows, apps, Dataverse and Power Pages, driven in plain language.
An MCP server that lets an agent see the screen and control macOS, Windows and Linux: screenshots, clicks, OCR and screen memory
DesktopManager gives an agent Windows windows and monitors over MCP: it starts read-only and changes need separate flags.
README in short
It covers running in Docker, HTTPS through Caddy or a custom certificate, the web chat and persistent settings. Configs for Claude Code, Cursor, VS Code, Windsurf and Claude Desktop follow, then a table of 62 tools with requirements and notes on TLS.