Document automation, shipped before lunch.
Klaaro reads, analyses and extracts validated data from your documents while easily integrating in your infrastructure through API, SDK and MCP.
import { KlaaroClient } from "klaaro"// Authenticate with your API keyconst apiKey = "KLAARO_API_KEY"const klaaro = new KlaaroClient({ apiKey })// Upload a document by URLconst datasetId = "YOUR_DATASET_ID"const url = "https://example.com/invoice.pdf"const doc = await klaaro.uploadDocument({ datasetId, url })// Read extracted recordsconst { records } = await klaaro.getDocumentRecords(doc.id)You could build a document automation. You shouldn't.
OCR, layout parsing, classification, schema generation, extraction, human review, exports. Weeks of work to maintain. Klaaro ships it as one API call.
Build your own
- Stitch together OCR, a vision model, an LLM, and a schema layer
- Prompt-tune extraction per document type — then babysit it as layouts change
- Build queues, retries, pagination, webhooks, and a review UI yourself
- Own the GDPR, hosting and data-residency compliance burden
Klaaro as your ADP platform
- One upload call runs OCR → classify → schema → extract, fully managed
- Auto-generated schemas adapt when layouts change — no re-prompting
- Typed SDKs, cursor pagination, webhooks and human-in-the-loop review built in
- EU-hosted and GDPR-compliant by default — nothing to configure
Integrate Klaaro into your stack.
Call it from application code, script it, or hand it to your coding agent.
The v1 API
A clean, cursor-paginated REST surface at /api/v1. Upload, poll, read records, export, and wire webhooks — full control over every HTTP call.
Official SDKs
Thin, typed wrappers — no hidden retries, no magic. Auth, multipart uploads, structured errors, pagination and webhook verification handled. Methods map 1:1 to routes.
MCP server
Give Cursor, Claude Code or Copilot structured access to your datasets and records over Streamable HTTP. OAuth in your editor, API keys for CI.
Every document runs through the same pipeline.
Klaaro runs highly parallel. Re-run from any stage if needed.
Ingest
Any file type, from your DMS, Drive, mailbox, SDK, API or MCP.
Parse & OCR
Layout understanding, handwriting, segmentation and model routing.
Classify
Assign each document to one or more of your dataset's classes.
Schema
Auto-generate an extraction schema for every new class it sees.
Extract
Validated records, ready to read via API, review, or export.
Schema generated with AI
Depending on the classification, Klaaro creates a schema relevant for that specific document type and, based on the information in your file, hands you a zero-shot configuration.
Change schemas with confidence
Define eval sets from your approved records and run branch evaluations every time you touch a schema. Measure extraction accuracy before you ship — for automations that stay robust as they evolve.
Klaaro/Evaluation
Evaluation(128 documents)
Manage approved-record eval sets and run branch evaluations.
Run history
| Status | Commit | Branch | Match | Progress | Duration | When |
|---|---|---|---|---|---|---|
| a3f8c21 | feat/vat-fix | 98.4% | 2m 14s | 2h ago | ||
| c91b0e4 | main | 96.1% | 1m 52s | 5h ago | ||
| f2d47a1 | feat/line-items | 88.7% | 3m 03s | 1d ago |
Enterprise-ready security meeting European standards.
Compliance your security team will sign off on.
GDPR compliant
Fully compliant with GDPR. Data-processing agreements available on request.
Hosted in the EU
All processing and storage happens on EU servers. Your data never leaves Europe.
No training on your data
Your documents are never used to train AI models. Your data stays exclusively yours.
Enterprise-grade
End-to-end encryption, scoped API keys, role-based access and audit logs.
