Skip to content

Harness MCP Server

Access Harness platform data for CI/CD, feature flags, and cloud cost management.

Capabilities
CI/CD Feature Flags Cloud Cost Deployments
Maintained by
Harness
Hosting
Hosted endpoint, or run it locally
Authentication
OAuth sign-in
Last checked

Hosted endpoint must be enabled for your account by Harness Support.

How to connect the Harness MCP server

Pick your AI client and paste the snippet. Swap the YOUR_… placeholders for your own values. New to this? Our guide to how MCP clients and servers fit together explains what each piece does.

Hosted server

Claude Code

Terminal
claude mcp add --transport http harness https://mcp.harness.io/mcp

Then run /mcp inside Claude Code and choose Authenticate to sign in.

Full Claude Code setup guide →

Claude Desktop and claude.ai

Settings → Connectors → Add custom connector
Name: Harness
Remote MCP server URL: https://mcp.harness.io/mcp

Custom connectors need a Pro, Max, Team or Enterprise plan. On Team and Enterprise an owner adds them in admin settings.

Full Claude Desktop setup guide →

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project)
{
  "mcpServers": {
    "harness": {
      "url": "https://mcp.harness.io/mcp"
    }
  }
}
Full Cursor setup guide →

VS Code

.vscode/mcp.json
{
  "servers": {
    "harness": {
      "type": "http",
      "url": "https://mcp.harness.io/mcp"
    }
  }
}
Full VS Code setup guide →

ChatGPT

Business, Enterprise and Edu plans (and Pro, read-only) can add this endpoint as an app in developer mode. ChatGPT setup guide →

+ Run it locally instead

Claude Code

Terminal
claude mcp add --env HARNESS_API_KEY=YOUR_API_KEY --transport stdio harness \
  -- npx -y harness-mcp-v2@latest
Full Claude Code setup guide →

Claude Desktop

claude_desktop_config.json (Settings → Developer → Edit Config)
{
  "mcpServers": {
    "harness": {
      "command": "npx",
      "args": [
        "-y",
        "harness-mcp-v2@latest"
      ],
      "env": {
        "HARNESS_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Fully quit and reopen Claude Desktop after saving.

Full Claude Desktop setup guide →

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project)
{
  "mcpServers": {
    "harness": {
      "command": "npx",
      "args": [
        "-y",
        "harness-mcp-v2@latest"
      ],
      "env": {
        "HARNESS_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}
Full Cursor setup guide →

VS Code

.vscode/mcp.json
{
  "servers": {
    "harness": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "harness-mcp-v2@latest"
      ],
      "env": {
        "HARNESS_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}
Full VS Code setup guide →

Frequently asked questions

What is the Harness MCP server?

It is a Model Context Protocol (MCP) server — a small piece of software that gives AI assistants like Claude a secure, governed connection to Harness. Access Harness platform data for CI/CD, feature flags, and cloud cost management.

Is there an official Harness MCP server?

Yes. The Harness MCP server is an official integration, maintained by the vendor or the Model Context Protocol project, which makes it the safest default for business use. It is maintained by Harness.

Is the Harness MCP server hosted, or do I run it myself?

Both. Harness runs a hosted endpoint at https://mcp.harness.io/mcp, so most people just paste that URL into their AI client. You can also run the server locally with npx, supplying your own credentials.

Do I need an API key for the Harness MCP server?

No API key is needed for the hosted server. You sign in through Harness's own OAuth screen, and the assistant only gets the access your account has.

What can an AI assistant actually do with Harness?

Through this server an assistant can work with ci/cd, feature flags, cloud cost, deployments — reading from and acting on Harness directly instead of you copy-pasting between windows. What you allow it to do is controlled by the permissions you grant.

How do I set this up for my business?

The setup section above has copy-paste config for Claude Code, Claude Desktop, Cursor and VS Code. Once it is running against real systems, the question becomes what catches it when it acts wrongly, which is what Crox's Safety Diagnostic looks at.

New to MCP? Start with our plain-English guide to the Model Context Protocol or see how to connect AI to your business tools without writing code .

Done-for-you integration

Already running this in production?

An agent with a live connection to your tools is a model that can act. Crox looks at what catches it when it acts wrongly: the layers around it, who can stop it, and where a near miss gets reported. Two days, one team, one model.