Dokyumi MCP integration

Connect your Dokyumi account from ChatGPT or another MCP client using OAuth. Server URL: https://dokyumi.com/mcp.

From a first document to a repeatable workflow

Browse 200 workflow starters with field definitions, review checks and labeled synthetic examples. Saving requires your agreement and the separate schema-write permission. It respects your saved-schema limit and reuses an existing schema without changing its fields. Browsing, saving, checking credits, searching and exporting spend no extraction credits.

The connection reads only the organization owned by the signed-in account. With extraction permission, it processes a selected attached PDF, JPEG, PNG, TIFF or WebP up to 20 MB using an active saved schema. An exact-byte inline fallback is limited to 32 KiB.

Credit preflight is an estimate from the page count you provide, not a reservation. Extraction checks the actual pages and current balance, stores the document and result, and uses one existing credit per up-to-five pages. Confirm credit use before extracting and reuse the same request key to retry. Review an interrupted extraction in your history before starting another request.

Search history by filename, schema or status, including results needing review. CSV and XLSX downloads support 1–50 selected completed or review-needed results and require signing into the same account in your browser. Exports include values, confidence, validation and links back to each result; nested values are preserved as JSON text. Oversized exports ask you to select fewer results.

Review confidence and validation errors before relying on data. The connector does not verify authenticity or provide legal, medical or tax determinations. No checkout or plan change occurs through MCP. The bounded synthetic reviewer uses its approved invoice fixture and existing schema; it does not validate general customer extraction accuracy.

Manage schemas · Review extractions · Privacy · Terms