Invoice data extraction API, for your code or your AI agent
Send invoices, receipts or bank statements, say in plain words what to extract, and get structured rows back as JSON, XLSX or CSV.
It is the same extraction as the web app, on the same account and credits, and everything it runs appears in your dashboard.
A model reading invoices on its own can hallucinate a value or skip a page, and nobody finds out. Here:
- A panel of AI agents has to agree on every value
- What the panel cannot agree on is flagged as Review Needed
- A page or document that fails is reported, never skipped in silence
- A single PDF up to 5,000 pages, a batch up to 6,000 files, every page and every file read and checked the same way as the first
- Where the documents leave something unsettled, it can stop and ask your agent instead of deciding on its own
Two ways to connect
Build it into code you keep, or have your agent run the extraction. Either way it is the same API key and the same credits.
For code
You build it into your code
An integration you write and keep, or have a coding assistant write for you. The same instructions produce the same columns and formats for every document, so the result imports without hand-fixing.
Each Copy docs button puts that path's complete docs on your clipboard, to paste into a coding assistant that builds the integration for you.
Node.js
Node.js 18 or later. One call uploads, submits, waits and returns the result; separate methods run each step.
npm install @invoicedataextraction/sdkPython
Python 3.9 or later. One call uploads, submits, waits and returns the result; separate methods run each step.
pip install invoicedataextraction-sdkREST API, any language
Everything the SDKs do is a handful of JSON endpoints under api.invoicedataextraction.com/v1. The reference carries every endpoint, field, limit and error, with a ready-to-run script.
For agents
Your agent runs the extraction
An agent that will run the extraction itself, for you, uses the skill or the MCP server. Both take it through the same steps: upload the files, submit, wait, answer what the extraction asks, read the result.
The skill
One file teaches an agent those steps as HTTP requests from a shell. Install it from this domain with the skills command-line tool, which places it for Claude Code, Codex, Cursor, OpenClaw, Hermes and other harnesses.
npx skills add https://invoicedataextraction.comThe MCP server
One remote server gives an agent the same steps as tools, in a harness that connects to MCP servers, Claude Code, Codex, Cursor, Hermes, OpenClaw and Gemini CLI among them. Your API key is the credential, and the docs carry the setting for each harness.
https://mcp.invoicedataextraction.com/mcpOr paste this to any agent
Nothing to install. Your agent reads the guide and takes it from there. When it needs an API key it tells you where a key is created: in the dashboard, on any account, with no card.
Read https://invoicedataextraction.com/docs/agents.md and set up invoice extraction for me.What it costs
You pay for pages processed and nothing else. There is no plan to size and no monthly charge: a quiet month costs nothing, and credits bought for a busy month keep for 18 months. One balance serves the web app and the API, and a person creates the API key in the dashboard.
For agents and tooling
The same docs in the forms a machine reads. Every docs page has a Markdown twin at its own address plus .md, and llms.txt indexes all of it.