@siddharthasingh/mcp-healthcheck
v0.1.0-alpha.0
Published
Trust-score scanner for MCP servers. Lighthouse for MCP.
Downloads
38
Readme
mcp-healthcheck
Trust-score scanner for MCP servers. Lighthouse for the Model Context Protocol.
Evaluate MCP servers on liveness, schema validity, security, freshness, tool quality, and documentation to output an objective, multi-dimensional trust score.
Features
- Scoring Engine: Calculates a 0-100 overall score based on 7 heavily-weighted categories.
- Deep Validation: Tests SSRF vulnerabilities, TLS configurations, and validates tool JSON schemas.
- Reporting: Exports detailed findings in JSON, HTML (with charts), and Markdown.
- Shields.io Badges: Instantly generates trust badges for your README.
- Continuous Monitoring: Run in background mode to record stability over time using SQLite.
Installation
npm install -g @siddharthasingh/mcp-healthcheckQuick Start
Scan an MCP server via Stdio:
mcp-healthcheck scan "npx -y @modelcontextprotocol/server-postgres"Scan an MCP server over HTTP/SSE:
mcp-healthcheck scan https://example.com/mcpGenerate a visually rich HTML report:
mcp-healthcheck report --format html > report.htmlGet a Markdown badge for your README:
mcp-healthcheck badge <server-id>Scoring Categories
The trust score is a weighted sum of the following:
- Liveness (20%): Handshake time, connection success, response latency (P50/P95).
- Schema Validity (15%): Adherence to the official JSON Schema spec.
- Security (25%): TLS verification, SSRF protections, Secrets leakage.
- Freshness (10%): Repository activity and commit recency.
- Tool Quality (15%): Descriptions, example usages, error handling.
- Docs (10%): Presence of
.well-known/mcp.jsonand install instructions. - Behavior (5%): Rate limits, idempotency.
License
MIT
