MeterSphere

MeterSphere MCP Server implements the Model Context Protocol to expose testing tools and resources over standard MCP channels. It lets you interact with testing services, fetch interface definitions, run test cases, and receive results directly from your preferred MCP client, such as IDEs or AI assistants, speeding up feedback and aligning development with testing.

How to use

You connect an MCP client to the MeterSphere MCP Server using either an HTTP/SSE channel or a local stdio/docker channel. Once connected, you can discover available testing tools, fetch interface definitions, execute test cases, create mock services, and perform compatibility checks directly from your editor or assistant. This enables testing workflows to move left into development and accelerates feedback without switching tools.

How to install

Prerequisites: you need a Java development environment and Maven to build the server.

# Clone the MCP server repository
git clone https://github.com/metersphere/metersphere-mcp-server.git
cd metersphere-mcp-server

# Build the server
mvn clean install

Configuration and deployment

You can connect using two MCP client methods described here. First, expose an HTTP/SSE endpoint for clients to connect. Second, run a local stdio-based server via Docker for development or local testing.

# Example: expose an HTTP/SSE endpoint (standard server URL/port used by the client)
# This is the runtime deployment step; exact server start commands depend on your environment.
# The following is a placeholder for the typical deployment approach.

# Start the server (using your preferred deployment method)
# java -jar metersphere-mcp-server.jar --server.port=8000

Security and permissions

The server supports AK/SK authentication to secure access. Role-based permissions ensure only authorized roles can call sensitive operations such as retrieving or processing project data.

Client connection examples

The following client configurations show practical ways to connect to the MCP server from common clients. Each snippet demonstrates how to establish a connection and supply required credentials.

Appendix: example client configurations

The server can be connected via HTTP/SSE or via a local Docker-based stdio channel. The configurations below reflect the two common approaches demonstrated in practical setups.

# VS Code - SSE connection (http/sse)
{
  "servers": {
    "ms-mcp-server-sse": {
      "type": "sse",
      "url": "http://localhost:8000/sse",
      "headers": {
        "accessKey": "your-access-key",
        "signature": "your-signature",
        "meterSphereUrl": "http://your-metersphere-url"
      }
    },
    
    # Docker-based stdio connection
    "ms-mcp-server-docker": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run", 
        "-i", 
        "--rm", 
        "-p", 
        "8000:8000", 
        "-e", "accessKey=${env:accessKey}",
        "-e", "signature=${env:signature}",
        "-e", "meterSphereUrl=${env:meterSphereUrl}",
        "metersphere-mcp-server"
      ],
      "env": {
        "accessKey": "your-access-key", 
        "signature": "your-signature",
        "meterSphereUrl": "http://your-metersphere-url"
      }
    }
  }
}

What you need to know about running modes

HTTP/SSE mode provides a remote endpoint you can call from any MCP-compatible client. Docker-based stdio mode lets you run the server locally in a container and connect via a standard input/output stream.