extract_invoiceLABSCO SUMMARY
An invoice to a bookkeeping entry in three calls.
Each step feeds the next: extraction returns an object, validation takes that same object and recalculates the totals, and categorisation turns the vendor and line items into an account code — which is the whole manual part of processing a bill.
What it is
An MCP service for invoice processing, served over Streamable HTTP and paid per call with x402 on Base mainnet. Three tools: extract structured fields from an invoice, validate an extracted invoice, and classify it into an accounting expense category. The same three operations are also exposed as a REST API.
What you get
- Vendor, client, invoice number, dates, line items, subtotal, VAT, total, bank details, payment terms and a confidence score
- Input accepted as pasted text, a public URL, or a PDF encoded as base64
- A completeness and arithmetic check returning severity-ranked issues, recalculated totals and a totals match flag
- An expense category with a confidence score, a suggested nominal account code and a tax-deductible likelihood
- Fourteen fixed categories, from software subscription and cloud infrastructure through to payroll and unknown
Requirements
You run it yourself: after npm install, npm run dev serves it at http://localhost:3101/mcp, the address the configuration example points at. Payments stay off until you enable them and set a receiving address, both through environment variables. Each of the three operations carries its own per-call price.
Setup effort
Build from source — git clone https://github.com/dannydaley76/invoice-intelligence-mcp.git
USE CASES
01
Getting the numbers off an invoice PDF
Send the PDF as base64, a public URL or pasted text, and get the fields back as structured data with a confidence score.
02
Catching an invoice whose totals do not add up
Validation recalculates subtotal, VAT and total and returns a totals match flag with severity-ranked issues.
03
Coding a bill to the right account
Vendor, description and line items become a category, a nominal code and a tax-deductible likelihood.
04
Working through a month of bills
Extract, validate and categorise in sequence, so what lands in the ledger has already been checked.
3 tools
Listed in project docs. Your client may show a different list.
validate_invoicecategorise_expenseBEFORE YOU INSTALL
What to know first
Tools you get
3 tools named in the project's documentation.
Requires
- Node.js and npm to run the server
- an MCP client that connects over Streamable HTTP
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
This one is built from source rather than installed with a single command.
Server setup
TERMINAL
git clone https://github.com/dannydaley76/invoice-intelligence-mcp.git
cd invoice-intelligence-mcp
npm install
cp .env.example .env
npm run dev # http://localhost:3101MCP Clients
Any MCP client — Local
With npm run dev running, point a client that supports Streamable HTTP at the local endpoint. Payments are off by default (PAYMENTS_ENABLED=false in .env).
ANY MCP CLIENT — LOCAL · JSON
{
"mcpServers": {
"invoice-intelligence": {
"url": "http://localhost:3101/mcp",
"type": "http"
}
}
}
