Skip to main content

MCP Server

Integrate Rigour with AI agents via Model Context Protocol (MCP).

Overview​

Rigour provides an MCP server that allows AI coding agents to validate their changes before finalizing a task. Unlike traditional linters, Rigour is designed to be consumed by agents to enforce engineering standards.

Installation​

npx @rigour-labs/mcp

Or install globally:

npm install -g @rigour-labs/mcp

Integration Cookbook: Client Recipes​

Rigour can be integrated with any MCP-compliant client. Below are optimized configuration recipes for the most popular AI agentic tools.

🤖 Claude Code​

Claude Code consumes MCP servers via its internal config.

# Add Rigour to Claude Code
claude mcp add rigour npx -y @rigour-labs/mcp

🖱️ Cursor​

Cursor can be configured globally or per-project.

Global setup:

  1. Open Cursor Settings > Features > MCP.
  2. Click + Add New MCP Server.
  3. Name: Rigour
  4. Type: command
  5. Command: npx -y @rigour-labs/mcp

Project-specific setup: Add this to your .cursor/mcp.json:

{
  "mcpServers": {
    "rigour": {
      "command": "npx",
      "args": ["-y", "@rigour-labs/mcp"]
    }
  }
}

🛠️ Cline / Roo Code​

Cline uses a dedicated settings file. Add this to your cline_mcp_settings.json:

{
  "mcpServers": {
    "rigour": {
      "command": "npx",
      "args": ["-y", "@rigour-labs/mcp"]
    }
  }
}

🌀 Antigravity​

If you are using Antigravity, Rigour is often pre-configured or can be added via the tool-call interface by pointing to the binary:

{
  "command": "npx",
  "args": ["-y", "@rigour-labs/mcp"]
}

🛰️ Gemini CLI / Firebase Genkit​

For programmatic usage or CLI wrappers that support MCP:

import { McpClient } from '@modelcontextprotocol/sdk';

const client = new McpClient({
  command: 'npx',
  args: ['-y', '@rigour-labs/mcp']
});

Available Tools​

The MCP server exposes tools for validation and project reasoning.

rigour_check​

Runs the full suite of Rigour quality gates on the current project state.

Arguments:

  • cwd (optional): Absolute path to the project root.

rigour_explain​

Provides a human-readable explanation of why the last quality gate check failed.

Arguments:

  • cwd (optional): Absolute path to the project root.

rigour_status​

A lightweight PASS/FAIL check that returns structured JSON. Ideal for agent polling.

Arguments:

  • cwd (optional): Absolute path to the project root.

rigour_get_fix_packet​

Retrieves the prioritized Fix Packet (Diagnostic) containing actionable instructions for the agent to resolve quality gate violations.

rigour_list_gates & rigour_get_config​

Utility tools that allow the agent to inspect active gates and the project's rigour.yml configuration.

The Stateless Workflow​

Rigour's MCP tools operate on the current filesystem state. Agents should follow this loop:

  1. Work: Apply changes to the codebase.
  2. Audit: Call rigour_check or rigour_status.
  3. Refine: If failures occur, call rigour_get_fix_packet or rigour_explain once to get instructions.
  4. Repeat: Resolve the issues and audit again until PASS.

Best Practices​

  1. Audit Before Done: Never claim a task is complete without a PASS status from Rigour.
  2. Read the Packet: Use rigour_get_fix_packet to understand how to refactor (e.g., extracting functions to reduce complexity).
  3. Reason about Rules: Use rigour_get_config to understand project-specific constraints like protected paths or forbidden dependencies.