API Documentation v1.1
Send a financial document, get its data back as structured JSON, and receive a webhook when processing finishes.
What do you need to do?
| You need to... | Start here |
|---|---|
| Send your first document and read the JSON back | Quickstart |
| Pull specific fields out of a financial document | Extraction |
| Know whether a document has been tampered with | Fraud analysis |
| Route an incoming file to the right process | Classification |
| Split a PDF that holds several documents | Segmentation |
| Check a printed address or company number against a reference | Field enrichment |
| Express your own business rules over extracted data | HoloLang |
| Collect a whole file of documents against one dossier | Dossiers |
| Be told when processing finishes, without polling | Webhooks |
| Know what is safe to retry, and which limits stop a call | Reliability and limits |
| Track consumption and processing volumes | Usage and Metrics |
| Run Holofin on behalf of client organizations | Sub-organizations |
Start building
Three ways in. Pick the one that matches how you work.
REST API
One authenticated POST sends a document through the whole pipeline. No SDK required.
Dashboard
Compose the workflow visually, test it on real documents, then trigger the same one by API.
AI agents
Point a coding agent at our MCP server, the installable skill, or the machine-readable docs index.
Everything starts with an API key from your developer settings. Keys are workspace-scoped and stay active even if the person who created them leaves the workspace.
The API is described by an OpenAPI 3.1 document, which you can feed to a client generator, import into Postman, or hand to an agent alongside the For LLMs files at the top of this page.
Workflows
Everything in Holofin runs through workflows. A workflow is a document processing pipeline that classifies, segments and extracts data from a document in a single API call.
Configure your workflow once in the dashboard, then trigger it via the API. The workflow handles routing, error handling and the aggregation of results from each step.