Product
Velrim is in open beta
Velrim is now generally available. Read the announcement
Velrim is a document extraction API. You send a JSON Schema (or a Zod or Pydantic model through the SDKs) and a document; you get back a typed object with a per-field state, a confidence score and page anchors that point at where in the document each value came from.
The free grant is about 500 pages and needs no card; it lasts 90 days once you claim it. Sign up to start.
What is live
The API: POST /v1/extract for synchronous extraction, async jobs with webhooks, and uploads for files up to 50 MB. Errors come in one shape with stable codes, and a failed extraction is not billed.
The dashboard: API keys, job history and a prepaid wallet.
SDKs: @velrim/sdk for TypeScript (Zod model in, parsed typed object out) and velrim for Python (Pydantic model in, validated instance out). Both also accept a plain JSON Schema object in place of a model, so neither library is required.
An MCP server at mcp.velrim.com/mcp, so agents can call velrim_extract directly.
Docs at velrim.com/docs, with markdown mirrors (docs.md, llms.txt) for agents.
Scores you can check
Every field comes back with a calibrated confidence score, and we publish the data behind the calibration at /reliability: reliability diagrams, ECE, Brier, AUROC and risk–coverage per document class, versioned by calibrator and measured on frozen splits. If the scores drift, that page is where it shows.
Pricing
Flat $0.02 per successfully extracted page, drawn from a prepaid balance you top up from $5. The balance never expires, and nothing recurs unless you turn on auto-recharge.
If you are extracting data from documents, start free. Questions to hello@velrim.com.