mcp

The Harness MCP Server exposes a collection of Microsoft Context Protocol (MCP) toolsets that let you automate and interact with Harness services from your tooling. It includes default tooling for core operations like listing and inspecting connectors, pipelines, dashboards, and secrets, as well as specialized toolsets for pipelines, pull requests, services, environments, infrastructure, connectors, registries, dashboards, and more. By connecting through MCP, you can drive Harness-driven automation, fetch execution URLs for pipelines, retrieve dashboard data, and orchestrate complex workflows using a consistent, tool-agnostic interface.

Typical usage involves starting the MCP server (via the provided Docker image in this configuration) and sending MCP requests to invoke specific toolset commands. Each toolset is named (for example, default, pipelines, connectors, secrets, etc.) and contains a defined set of operations such as list_, get_, and create_ actions. This enables you to compose automation that can query Harness resources, trigger runs, fetch results, and monitor state across multiple services from a single entry point.

Prerequisites:

• Docker installed and running
• Basic familiarity with MCP clients or tooling that can communicate with MCP servers

Install and run the MCP server using Docker:

1. Pull and run the Docker image:

   docker run -d --name harness-mcp-server -p 8080:8080 -e MCP_LOG_LEVEL=info harness/mcp-server:latest

2. Verify the server is responding (adjust port if you mapped differently):

   curl http://localhost:8080/v1/health

3. (Optional) Configure your MCP client to point to http://localhost:8080 and begin issuing MCP requests using the supported toolsets.

If you prefer building from source, ensure you have Docker and your preferred build tools installed, then follow the project-specific build instructions in the repository (often involving a containerized build or a local Node/Python environment, depending on the project).

Tips and common issues:

• Ensure the Docker image harness/mcp-server:latest is accessible in your environment (pull if necessary).
• If you customize environment variables, expose or map the required ports and verify connectivity from your MCP client to the server.
• The MCP server may expose multiple toolsets; refer to the server's API docs or help endpoints (e.g., /v1/help) to discover available commands.
• For debugging, set MCP_LOG_LEVEL to debug to obtain verbose logs.
• If authentication is required by your Harness instance, configure the appropriate auth tokens or credentials in the client requests; the server may support env-based or header-based authentication depending on deployment.
• When running behind a proxy or firewall, ensure the MCP server port is accessible and that MCP clients can reach it directly.