mcp-checkpoint

MCP Checkpoint is a security scanner for Model Context Protocol (MCP) servers. It automatically discovers MCP configurations, inventories tools, resources, and prompts from integrated MCP servers, and performs security analyses to detect issues such as prompt injections, tool shadowing, and resource tampering. The tool also provides baseline drift detection to identify changes against approved configurations and generates comprehensive reports in JSON or Markdown formats. You can use it to obtain an audit trail of configurations and findings, making it easier to secure MCP-based workflows across agents and IDE integrations.

To use the server, install the Python package, run the CLI, and leverage the scan and inspect commands. The scan command automatically detects configurations (and baseline.json if present) and performs security checks, while the inspect command helps you create or inspect a baseline. You can point the scanner at specific MCP configuration files with the --config option, generate a markdown report with --report-type md, and save outputs to a file with --output. The tool is designed to work with various MCP configurations and is compatible with common MCP server setups across Agentic IDEs and clients.

Prerequisites:

• Python 3.8+ installed on your system
• pip (Python package installer) available in PATH

Installation steps:

1. Create and activate a virtual environment (optional but recommended):

   • python -m venv venv
   • source venv/bin/activate # On Windows use: venv\Scripts\activate

2. Install MCP Checkpoint from PyPI:

   • pip install mcp-checkpoint

3. Verify installation:

   • mcp-checkpoint --help

4. Run an initial scan using default configuration discovery:

   • mcp-checkpoint scan

Note: You can also inspect MCP configurations and generate baselines or reports by using the inspect and scan commands with additional options as described in the help output.

Tips and considerations:

• The tool detects and analyzes MCP configurations integrated with major Agentic IDEs and clients automatically during scans.
• Runtime environment can impact discovery; ensure MCP servers are reachable and configured for local testing.
• Baseline management: use inspect to create a baseline file (baseline.json by default) and use that baseline in subsequent scans for drift detection.
• Reporting: you can generate Markdown reports for human-readable reviews or JSON reports for programmatic consumption via --report-type md or --report-type json (default).
• Logging: scan results are kept in logs and can be reviewed if you enable verbose or show-logs modes (via the CLI help).
• If you encounter issues with configuration discovery, verify that MCP config files are valid JSON and that network access to MCP servers is not blocked by firewalls or proxies.