The refinery API — for people and AI agents
Every dataset on this site is queryable as clean JSON. Every record carries
quality_score, verified_against (the independent authority), and
source_url (a re-queryable citation).
Endpoints
GET /datasets the catalog + filter vocabulary
GET /datasets/{ds}/records?city=&zip=&q= filtered query (see per-dataset filters below)
GET /datasets/{ds}/records/{record_id} one record
GET /datasets/{ds}/certificate the DAMA data-quality certificate
GET /docs interactive OpenAPI docs
Examples
curl 'https://toolbox2026.com/datasets/foreclosure/records?county=bexar&date_from=2026-07-01&limit=5' curl 'https://toolbox2026.com/datasets/childcare/records?city=Parker&min_capacity=100' curl 'https://toolbox2026.com/datasets/recalls/records?q=laceration&date_from=2026-01-01'
For AI agents (MCP)
Attach the refinery as an MCP server and query it with cited answers:
claude mcp add refinery -- python3 /path/to/refinery/serve/mcp_server.py
Or read /llms.txt for a machine-readable summary of everything here.
Datasets
city, zip, type, name, min_quality, min_capacitycounty, state, city, zip, type, date_from, date_to, min_quality, urgency, newly_listedfirm, product, recall_number, date_from, date_to, min_qualityagency, recipient, recipient_uei, award_type, min_amount, date_from, date_toAccess tiers & terms (v1, 2026-07-06)
Viewing this site is free and always will be. Anonymous API use is free up
to 25 records/request and 60 requests/hour — enough to build and test against. Higher volumes use
an API key: see pricing. Bulk artifacts (feed.csv, api/foreclosures.json)
are free during beta; keys will be required for bulk later — that change will be announced here
with a version bump, not silently. Journalists, researchers, and NGOs get free at-scale access —
apply here.
Data ethics: we sell property-level and aggregated public-record intelligence only. No
owner-contact appends, no person-targeting products, no homeowner lead lists. Every record cites
the public authority it came from.