8,535 66,496: how we indexed every MCP server on GitHub, npm and PyPI (and deleted 2,910 fake install counts)

Description and use

The MCP registry removed 2,910 fake install counts and corrected inflated GitHub stars, delivering accurate data for 66,496 servers and offering free search API and badges for maintainers.

Problem

The registry contained inaccurate data—invented install counts and inflated GitHub stars—making it unreliable for users and maintainers.

Outcome

The MCP registry now contains accurate, verified data for 66,496 servers, with install‑risk classification and trust scoring, and provides free search API and badge system for maintainers.

Workflow

• 1. Run multi‑registry indexing: GitHub code search, npm registry API scan, PyPI metadata extraction.
• 2. Deduplicate by repository identity across sources.
• 3. Verify GitHub data via API (stars, issues, last push).
• 4. Remove legacy rows with untraceable install counts (2,910 entries).
• 5. Re‑calculate GitHub stars from live API, discarding inflated values.
• 6. Re‑run certifications and purge stale data layer.
• 7. Publish transparency page and submission form.
• 8. Offer CLI tools (search, install‑risk, fingerprinting, credential verification).
• 9. Notify maintainers via opened issues with badge links.

Evidence

2,910 invented install counts — legacy listing rows whose install numbers could not be traced to any registry. All deleted.

Source

~2x inflated GitHub stars — catalog stars came from an old scrape; live API verification showed roughly double the real numbers.

Source