Claw.

The Planning Inspectorate publishes appeal decisions as PDF letters. Claw starts with an ingestion pipeline that collects those letters, extracts the text, and adds them to a search index that now covers more than 169,000 decisions and keeps itself current as new ones are published.

On top of that sits a full-text search engine with filters for decision type, local authority and date range. Results come back in under a second, and the original decision letter opens in the browser alongside them.

The AI research tools are grounded in the documents. A user opens any decision letter, asks questions in plain English, and gets answers drawn from the text of that letter rather than from a model's general knowledge. Around all of this is a complete commercial product, with team accounts, shared bookmarks and subscription billing.

What we built

• An ingestion pipeline that extracts text from PDFs and keeps the index current
• Full-text search with filters, returning results in under a second
• An in-browser viewer for the original decision letters
• AI question answering grounded in each document's own text
• Team accounts with shared bookmarks and libraries
• Subscription billing with a 14 day free trial