Skip to content

MCP_INTEGRATION.md

Latest commit

 

History

History
135 lines (100 loc) · 4.23 KB

MCP_INTEGRATION.md

File metadata and controls

135 lines (100 loc) · 4.23 KB

Poutine MCP Server Integration Guide

The Poutine MCP (Model Context Protocol) server allows AI coding assistants to analyze repositories and CI/CD pipelines for security vulnerabilities directly from your development environment.

Prerequisites

  1. Install Poutine: Follow the installation guide to install Poutine

  2. GitHub Authentication: Set up GitHub CLI authentication

    gh auth login

  3. Set GitHub Token Environment Variable: Before launching your AI coding assistant, export the GitHub token:

    export GH_TOKEN=$(gh auth token)

    The Poutine MCP server will automatically pick up the GH_TOKEN environment variable from your shell session.

Setup

Claude Code

claude mcp add poutine poutine mcp-server

Gemini CLI

gemini mcp add poutine poutine mcp-server

Other MCP-Compatible Clients

Add the following configuration to your MCP-compatible AI coding assistant:

"mcpServers": {
  "poutine": {
    "type": "stdio",
    "command": "poutine",
    "args": [
      "mcp-server"
    ],
  }
}

Note: The Poutine MCP server will automatically pick up the GH_TOKEN environment variable from your shell session. Make sure you've set it (see Prerequisites) before launching your AI coding assistant.

Available MCP Tools

Once configured, the following tools are available to your AI assistant:

analyze_org

Scan all repositories in a GitHub/GitLab organization.

Parameters:

  • org (required): Organization name
  • scm_provider (optional): "github" or "gitlab" (default: "github")
  • scm_base_url (optional): Base URL for self-hosted instances
  • threads (optional): Number of parallel threads (default: 2)
  • ignore_forks (optional): Skip forked repositories (default: false)

analyze_repo

Scan a specific repository.

Parameters:

  • repo (required): Repository name in format "org/repo"
  • scm_provider (optional): "github" or "gitlab" (default: "github")
  • scm_base_url (optional): Base URL for self-hosted instances
  • ref (optional): Git branch or commit to analyze (default: "HEAD")

analyze_local

Scan a local repository by file path.

Parameters:

  • path (required): Local file system path to the repository

analyze_repo_stale_branches

Scan repository branches for pull_request_target vulnerabilities.

Parameters:

  • repo (required): Repository name in format "org/repo"
  • scm_provider (optional): "github" or "gitlab" (default: "github")
  • scm_base_url (optional): Base URL for self-hosted instances
  • threads (optional): Number of parallel threads (default: 5)
  • expand (optional): Expand output to full format (default: false)
  • regex (optional): Regex pattern for workflow matching (default: "pull_request_target")

analyze_manifest

Analyze CI/CD pipeline manifest content for security issues.

Parameters:

  • content (required): The complete YAML manifest content
  • manifest_type (required): Type of manifest - "github-actions", "gitlab-ci", "azure-pipelines", or "tekton"

Note: This tool is automatically called when AI assistants generate or modify CI/CD workflows to ensure security best practices.

Example AI Assistant Prompts

Here are some example prompts you can use with your AI coding assistant:

Organization-wide scan:

Use Poutine to scan all repositories in the <your_org> organization

Single repository analysis:

Analyze the security of the repository actions/checkout

Local repository scan:

Scan the repository at /Users/me/projects/myapp for supply chain vulnerabilities

Stale branch analysis:

Check the myorg/myrepo repository for pull_request_target vulnerabilities in stale branches

Workflow generation with automatic security validation:

Create a GitHub Actions workflow that runs tests on pull requests

(The AI will automatically use analyze_manifest to validate the generated workflow)

Additional Resources