Skip to main content
This feature is currently in preview. If you have any questions, please contact us.
Chainloop remote Model Context Protocol (MCP) server is a new way to interact with Chainloop natively from your AI clients or agents. This allows you to perform complex queries, craft compliance reports, or implement advanced agentic workflows. All of this done securely, using the data already stored and signed in Chainloop. info

​
How to connect to the MCP server

If you are running your own instance of Chainloop platform, first, make sure to enable the feature by following this guide.
The MCP server is currently exposed as a streamable API) endpoint at https://mcp.app.chainloop.dev/mcp. For authentication, the MCP server supports both User and API token authentication methods.

​
Client Examples

Our server is been tested against the following MCP Clients: Claude Desktop, Cursor, Visual Studio Code, and Dagger. Below you will find some examples, if you can’t find yours, please refer to their documentation.
Although many clients already support remote MCP servers, most have known limitations. That’s why you’ll see some examples leveraging mcp-remote package. This is considered a workaround for the time being, our production stance will be to use clients that fully support MCPs SSE or Streamable HTTP transports, but at the moment of the writing and for testing purposes, we’ll show you examples using mcp-remote instead.Using mcp-remote is considered a workaround for the time being, our production-ready stance will be to use clients that fully support MCPs Streamable HTTP transports, but as the moment of the writing, and for testing purposes, we’ll show you examples using mcp-remote instead.

​
Claude Desktop

You can configure the MCP server in Claude Desktop please follow this example to find the configuration file but add the following MCP configuration instead. 
{
  "mcpServers": {
    "chainloop": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.app.chainloop.dev/mcp",
        "--header",
        "Authorization: Bearer ${CHAINLOOP_TOKEN}"
      ],
      "env": {
        "CHAINLOOP_TOKEN": "REDACTED"
      }
    }
  }
}
Once you enable it and restart the Claude, you will set the list of tools next to the prompt. info

​
Cursor

For Cursor, we need to do something funny with the header Authorization:${AUTH_HEADER} this is due to known limitation.
{
  "mcpServers": {
    "chainloop": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.app.chainloop.dev/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer REDACTED"
      }
    }
  }
}
info

​
Claude Code

Claude Code also supports MCP servers. You can configure the Chainloop MCP server using the claude CLI command: 
claude mcp add-json chainloop '{"command":"npx","args":["mcp-remote","https://mcp.app.chainloop.dev/mcp","--header","Authorization: Bearer ${CHAINLOOP_TOKEN}"],"env":{"CHAINLOOP_TOKEN":"REDACTED"}}'
Replace REDACTED with your Chainloop API token. Alternatively, you can manually add the configuration to your Claude Code settings file:
  1. Open Claude Code settings by running the command /settings or by opening the settings file directly at ~/.claude.json (Linux/macOS).
  2. Add the following MCP server configuration: 
{
  "mcpServers": {
    "chainloop": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.app.chainloop.dev/mcp",
        "--header",
        "Authorization: Bearer ${CHAINLOOP_TOKEN}"
      ],
      "env": {
        "CHAINLOOP_TOKEN": "REDACTED"
      }
    }
  }
}
  1. Replace REDACTED with your Chainloop API token.
  2. Restart Claude Code for the changes to take effect.
Once configured, Claude Code will automatically connect to the Chainloop MCP server, and you’ll be able to interact with Chainloop directly from your AI coding sessions.