Skip to content

Use with AI agents (MCP)

The Olaverse MCP server gives AI agents the API as tools. Ask your agent to “pull the total and date out of this receipt” or “which language is this message in?”, and it calls Olaverse directly. It works with any client that supports the Model Context Protocol: Claude Desktop, Claude Code, Cursor, Windsurf and others.

Tool What it does
extract_document A document (text, a JPEG/PNG photo or scan, or a PDF) to JSON shaped like a template, with a calibrated confidence for every field. Long documents run as a job.
get_extraction_job Collects a long-document result that’s still running.
decide Typed questions about a message, ticket or record, with calibrated probabilities.
detect_language The language of a text, among 608.
restore_diacritics Tone marks, accents and Arabic vowel marks back in 11 languages.
generate_questions Search-style questions a passage answers, in its own language.

You need an API key from the developer console. Connect to the hosted server (nothing to install), or run it locally with uv (uvx downloads it the first time; the local server can also read files on your machine).

Point any client that supports remote MCP servers (Claude, ChatGPT and Cursor connectors, Arcade, Smithery and others) at:

https://mcp.olaverselab.com/mcp

and send your key as a header:

Authorization: Bearer your-key

The hosted server takes documents as text, or as an image or pdf in base64. It doesn’t read files or fetch URLs.

Restart (or reconnect) the client, and the six tools appear.

Ask in plain words:

  • “Extract the merchant, date, total and line items from ~/Downloads/receipt.jpg.”
  • “Read these MoMo messages and give me the amount, recipient and reference of each.”
  • “Is this customer asking for a refund, and how urgent is it?”
  • “Which language is ‘Ina kwana’? Then add the tone marks back to ‘se eranko naa si gbo o’.”

Every call is billed to your key at the usual prices, exactly like a direct API call.

Variable Default
OLAVERSE_API_KEY required Your API key.
OLAVERSE_JOB_WAIT 900 Seconds a long-document extraction waits before handing back a job id for get_extraction_job.

Scanned PDFs without a text layer aren’t read directly: send their pages as JPEG or PNG images.

The server is open source (Apache-2.0): github.com/Olaverse-Labs/olaverse-mcp.