talonic_get_document

Fetch full metadata for a single document by ID, including processing log and link URLs.

`talonic_get_document` retrieves comprehensive information about a document that has already been ingested into your Talonic workspace. This includes the file's status, page count, size, detected document type and language, the source that uploaded it, and direct links to the document in the Talonic dashboard and API.

This tool is useful when the agent needs to check whether a document has finished processing, verify its detected type, or retrieve the dashboard link so the user can view the original. It does not return extracted data — for that, use `talonic_extract` with the same document_id.

The links.dashboard URL in the response is particularly valuable for human-in-the-loop workflows. When the agent needs the user to verify an extraction against the original document, it can present this link as a direct way to view the source in the Talonic web interface.

Example

{
  "document_id": "doc_8f3a1b2c-4d5e-6f7a-8b9c-0d1e2f3a4b5c"
}

{
  "id": "doc_8f3a1b2c-4d5e-6f7a-8b9c-0d1e2f3a4b5c",
  "filename": "invoice-2026-001.pdf",
  "status": "completed",
  "pages": 2,
  "size_bytes": 184320,
  "mime_type": "application/pdf",
  "type_detected": "Invoice",
  "language_detected": "de",
  "source": { "id": "src_a1b2...", "type": "api" },
  "links": {
    "self": "https://api.talonic.com/v1/documents/doc_8f3a...",
    "extractions": "https://api.talonic.com/v1/documents/doc_8f3a.../extractions",
    "dashboard": "https://app.talonic.com/documents/doc_8f3a..."
  }
}

Example: human-in-the-loop verification

// After extracting data with low confidence on a critical field:
// Agent calls talonic_get_document to get the dashboard link:
{
  "document_id": "doc_8f3a1b2c-4d5e-6f7a-8b9c-0d1e2f3a4b5c"
}

// Agent presents to user:
// "The total_amount field has confidence 0.68.
//  Please verify against the original document:
//  https://app.talonic.com/documents/doc_8f3a..."

The status field in the response indicates the document's processing state. Common values are completed (ready for extraction and markdown retrieval), processing (still being ingested), and failed (ingestion encountered an error). Agents should check the status before attempting follow-up operations — calling talonic_extract on a document that is still processing will either wait or fail depending on the API's current behaviour.

The source object in the response tells you how the document was uploaded. A type of api means it was uploaded via the MCP server or REST API, while dashboard means it was uploaded through the Talonic web interface. The source.id can be useful for filtering documents by upload source when a workspace has documents from multiple integration points.

This tool is free to call — it does not consume extraction credits. Agents can call it as often as needed to check document status, retrieve links, or verify metadata without any cost concern. This makes it safe for polling scenarios where an agent needs to wait for a document to finish processing before proceeding with extraction.