Finance · AP · Ready to use

Invoice Processor

Production multi-agent AP built with Pydantic AI — extract invoice data from PDFs or email, match purchase orders, flag discrepancies, and recommend Approve / Needs Review / Reject with human-in-the-loop.

▸ Extractor · Matcher · Reviewer · Orchestrator
▸ Strong Pydantic models · async pipeline · DI for Gmail / QuickBooks / NetSuite later
▸ Prefer model xai:grok-4.5 · works offline with rule fallbacks

Pydantic AI PO match HITL FastAPI

1. Extract

PDF / email / text → structured InvoiceData.

2. Match

PO lookup + amount/vendor/qty discrepancy rules.

3. Review

Approve, needs review, or reject — with HITL flags.

4. Prepare

Accounting stub ready for QuickBooks / NetSuite.

Deploy on this machine

Code: agents/invoice-processor · Guide: agents/invoice-processor/README.md

cd agents/invoice-processor
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
copy .env.example .env
python scripts/smoke_test.py
python scripts/run_example.py
python -m src.cli serve

API: http://localhost:8799 · Optional: set XAI_API_KEY for Grok specialists.

Safety: High-value invoices and material PO discrepancies force human review before payment prep.

Related agents

Pair with InvoiceForge for inbox watch, AR aging, and the visual review dashboard.

Back to marketplace See InvoiceForge