§ v. Specimen2026
Invoice Crawler.
Invoice Crawler is a desktop tool that extracts structured data from invoice images and PDFs — dates, totals, VAT, company names, and any other fields configured per firm — and exports the results as a clean CSV. V2 is a ground-up rebuild of a 2023 internal tool: the original used Tesseract OCR plus a multi-agent GPT writer/editor pattern to compensate for unreliable structured output, and modern vision APIs with schema enforcement collapse that whole pipeline into a single multimodal call per invoice — cheaper, faster, and considerably more accurate. Profiles per firm, encrypted API keys via the OS keychain, and a reviewable results table before export.
§ 00Downloads · v2.0.2
- macOS · Apple SiliconInvoice-Crawler-2.0.2-arm64.dmg123 MBGet ↓
- macOS · IntelInvoice-Crawler-2.0.2.dmg128 MBGet ↓
- Windowsinvoice-crawler-2.0.2-setup.exe99.6 MBGet ↓
- LinuxInvoice-Crawler-2.0.2.AppImage134 MBGet ↓
Older releases on GitHub.
§ 01Features
- 01Vision-based extraction — no OCR, no Tesseract, no Poppler
- 02Multi-language out of the box — any script, no per-language setup
- 03Cost-efficient — ~$1–3 per 1,000 invoices via Haiku 4.5 / GPT-4o-mini
- 04Configurable fields with synonyms and formatting hints
- 05Firm profiles — preset fields, AI provider, model, and key per firm
- 06Editable results before CSV export
- 07Batch processing with live progress and cancellation
- 08Local-first — keys encrypted in OS keychain, no cloud storage
§ 02Stack
- TypeScript
- Electron
- Anthropic Claude
- OpenAI
- React
Tags
Desktop·AI·OCR·Accountancy·Electron·Vision