Independent public-records infrastructure for California.
FOIA Gras builds a searchable, citable, AI-ready corpus of every publicly-posted board meeting and document from California government. The California project extends that to the agencies whose budgets, timelines, and outcomes matter most to people who actually live and build here.
The primary record of what California agencies do is their meeting packet: the agenda, the staff report, the contract, the change order, the audit response. Those packets are public — and yet reading them is a second job. They live on platform-specific portals, behind PDF scanners, inside 400-page attachments nobody opens. Accountability narrows to the few reporters, advocates, or staff who can afford the time.
Meanwhile the systems that produce those documents keep getting more expensive, more complicated, and more consequential. High speed rail. School budgets. Bond-funded construction. Enterprise payroll. The public record exists. The public can't read it.
We scrape, OCR, and index the full document history of each agency we cover — then expose it through three surfaces built on the same backend:
- Inspector General — an AI chat that answers questions with per-page citations back to the original PDF and meeting.
- MCP server — the same corpus, exposed to Claude Desktop, Cursor, ChatGPT, or any other Model Context Protocol-capable tool. Run your own research from your own workflow.
- Newsletter & investigations — long-form reporting when a pattern in the documents warrants it.
The entire pipeline — scraper, OCR, embedding, chat, MCP — is one codebase. That's the leverage: once an agency is in, every surface lights up at once.
Next: expanded California coverage including major city councils, transit agencies, and UC/CSU governance. Agency suggestions welcome.
FOIA Gras is built and run by Tom Hayden — a one-person platform covering scraping, OCR, vector search, AI agents, the chat UI, and the newsletter stack. California is production infrastructure, not a demo.
Before FOIA Gras: CTO at Engineering Director; data and fraud infrastructure at Grubhub and Facebook — including the pipelines behind both IPOs and $10M+ of fraud-detection impact. Adjunct faculty at Northwestern Medill. CS PhD coursework at Northwestern (left to join Facebook).
How it started: a local FOIA habit during COVID turned into the Evanston newsletter that now functions as the de facto inspector general for three governments. California is the same model at state scale — same codebase, same AI, aimed at the agencies where the dollars and the decisions are largest.
Investors, partners, and funders — please say hello. I read every email personally.
Accountability infrastructure
is not free, but it's cheap.
The California expansion — more agencies, deeper historical coverage, faster ingest, better AI — is funded by readers, civic supporters, and mission-aligned donors. Three ways to help: