reportportal

The ReportPortal MCP Server acts as a bridge between your ReportPortal instance and AI chat assistants, enabling natural-language queries about test runs and results. It follows the MCP guidelines and exposes commands that let AI tools fetch information such as recent runs, failure counts, and detailed error data directly from ReportPortal. When running locally in stdio mode, you can interact with the server via standard input/output; in HTTP mode, the server exposes an HTTP/SSE endpoint for remote multi-client access. To use it, configure the MCP client (your AI tool) to point at the server and provide the necessary RP_HOST, RP_PROJECT (optional), and an API token for authentication. Tools like Claude Desktop, Cursor, or GitHub Copilot can leverage these built-in commands to query runs, list results, or perform basic analyses within your chat or code editor workflow.

Prerequisites:

• Docker (recommended) or pre-built binaries for your OS
• Access credentials for your ReportPortal instance: RP_HOST, RP_PROJECT (optional), RP_API_TOKEN
• If using HTTP mode, MCP_MODE must be set to http

Option 1: Local installation via Docker (recommended)

1. Ensure Docker is installed and running.
2. Run the MCP server container with the required environment variables:

docker run -i --rm \
  -e RP_API_TOKEN=your-api-token \
  -e RP_HOST=https://your-reportportal-instance.com \
  -e RP_PROJECT=YourProjectInReportPortal \
  reportportal/mcp-server

3. The server will start in stdio mode by default. For HTTP mode, set MCP_MODE=http in the environment or as configuration when starting the container (see For developers in the repository).

Option 2: Local installation via pre-built binaries

1. Download the OS-specific pre-built binary from the official releases page.
2. Run the binary in stdio mode (default) or pass MCP_MODE=http to enable HTTP mode:

./reportportal-mcp-server stdio
# or for HTTP mode
MCP_MODE=http ./reportportal-mcp-server http

3. Provide configuration via environment variables as shown in the examples (RP_API_TOKEN, RP_HOST, RP_PROJECT).

Option 3: Connecting to a Remote MCP server If you are connecting to an already deployed MCP server over HTTP, configure the remote connection in your mcp.json as shown in the README. You do not run the server locally in this case; you simply point your AI tool at the remote URL and provide the required headers (Authorization with a Bearer token and optional X-Project).

Tips and notes:

• Always keep RP_API_TOKEN secure and do not commit it to version control.
• In stdio mode, AI tools interact via direct messaging streams; in HTTP mode, ensure the MCP server is deployed and reachable at a public or network-accessible URL.
• If RP_PROJECT is omitted, the server will use the project from each request, but providing it in the env helps defaults when not overridden.
• When using Docker, the container will pull reportportal/mcp-server from DockerHub; ensure you have network access to fetch the image.
• For remote usage, ensure MCP_MODE is set to http and the remote server is configured to accept connections from your AI tool.
• The MCP server supports common queries like listing runs, getting failure details, and summarizing test results; consult the tool’s docs for the exact MCP command set supported by this server.