Show HN: CharityVerify – Trust scores for 138K Canadian charities
buchler Tuesday, February 24, 2026
I built CharityVerify to make Canadian charity data actually usable.
The Canada Revenue Agency publishes T3010 forms for every registered charity, but they're scattered across clunky databases with no standardization or comparability. I collected 15 years of filings for all 138,203 charities and built a trust scoring system on top.
Stack: - Python + Playwright for CRA data collection (4s rate-limited) - PostgreSQL (Supabase) — 12 T3010 tables, 138K charities, 457K directors, 362K directorship links - Express.js REST API on Fly.io - Daily GitHub Actions sync for new filings - On-demand narrative generation via Claude Haiku
Scoring algorithm: Three 0-100 scores per charity: 1. Legitimacy (filing consistency, directorship stability, CRA compliance) 2. Effectiveness (program spending ratio, overhead, donation efficiency) 3. Compliance (sanctions screening, FATF risk, political activity limits)
Each charity gets a letter grade (A+ to F, or NR for insufficient data).
Findings: - Only 186 out of 85,507 registered charities scored A+ - Average effectiveness score: 51.6/100 - 487,692 flags generated (directorship overlap, compensation issues, filing gaps, etc.)
The core search/view is free. I'm building a tiered REST API for professional use cases (due diligence firms, grant-making orgs, etc.).
Code is closed-source for now, but the underlying CRA data is public domain. Happy to discuss the data pipeline, scoring methodology, or data collection approach.