# Talonic

> The data registry for unstructured documents. Agents extract once, query forever.
> Talonic ingests 25+ document formats and produces schema-validated structured data
> with per-cell confidence and source provenance. Built for AI agents and enterprise
> systems alike. EU-resident infrastructure, GDPR/HIPAA compliant, ISO 27001/42001
> aligned, DIN SPEC 91491 co-author.

## For agents

- [MCP Server](https://mcp.talonic.com/mcp): Hosted MCP, zero install. Connect from Claude, Cursor, or any MCP client with `Authorization: Bearer tlnc_...`.
- [Cost & balance headers](https://talonic.com/developers#billing): Every sync 200 returns `X-Talonic-Cost-Credits`, `X-Talonic-Cost-EUR`, `X-Talonic-Balance-Credits`. Track spend per call.
- [Credit balance & runway](https://talonic.com/developers#billing): `GET /v1/credits/balance` returns balance, 30-day burn rate, projected runway in days.
- [Auto top-up (human-gated)](https://talonic.com/developers#billing): Agents can call `POST /v1/billing/topup` with `billing` scope, but only humans can enable it via dashboard.
- [Idempotency](https://talonic.com/developers#auth): Pass `Idempotency-Key` on POST. 24h TTL, scoped per API key.
- [Sync vs async contract](https://talonic.com/openapi.json): ≤5 pages → 200 sync, larger → 202 async with `poll_url`.

## Three modes

- [Mode 1 — Extract everything](https://talonic.com/developers#trio): No schema. AI discovers every field. For prototyping or unknown docs.
- [Mode 2 — Extract a defined shape](https://talonic.com/developers#trio): Send schema, get exactly that shape. Three formats accepted: JSON Schema, simplified fields, flat key-type map.
- [Mode 3 — Query without re-extracting](https://talonic.com/developers#trio): Documents already ingested? Just filter and query. `POST /v1/documents/filter` returns structured rows. Zero AI calls, zero re-extraction.

## API quickstart

- [POST /v1/extract](https://talonic.com/docs/api/extract): Primary endpoint. multipart/form-data with `file` + optional `schema`.
- [GET /v1/extractions/:id](https://talonic.com/docs/api/extractions/get): Retrieve full extraction with confidence and provenance.
- [POST /v1/documents/filter](https://talonic.com/docs/api/search/filter): Mode 3 — query the registry.
- [Webhooks](https://talonic.com/docs/api/webhooks): HMAC-SHA256 signed events. Retry ladder 0s/30s/2m/8m/30m/2h/8h, then DLQ.
- [OpenAPI 3.1 spec](https://talonic.com/openapi.json): Machine-readable.

Auth: `Authorization: Bearer tlnc_live_...` on every request. Base URL: `https://api.talonic.com`.

Minimal example:
```
curl -X POST https://api.talonic.com/v1/extract \
  -H "Authorization: Bearer tlnc_live_..." \
  -F "file=@invoice.pdf"
```

## Concepts

- [Field Registry](https://talonic.com/docs/concepts/field-registry): Canonical fields that compound across every document.
- [Provenance & Confidence](https://talonic.com/docs/concepts/provenance): Every value points back to source region, with per-cell confidence (0.0–1.0).
- [4-Phase Pipeline](https://talonic.com/docs/concepts/pipeline): Resolve → Agent → Validate → Re-read. Confidence gate at 0.7 — protected cells cannot be overwritten.
- [Cases & Document Linking](https://talonic.com/docs/concepts/cases): Documents cluster automatically through shared identity/transaction/reference keys.
- [529-type ontology](https://talonic.com/docs/concepts/ontology): Language-agnostic document classification.

## SDKs

- [Node SDK](https://github.com/talonicdev/talonic-node): `npm install @talonic/node`
- [MCP Server](https://github.com/talonicdev/talonic-mcp): `npx -y @talonic/mcp@latest`

## Compliance & residency

- EU-resident infrastructure (Germany West Central)
- GDPR, HIPAA compliant
- ISO 27001 / ISO 42001 aligned
- DIN SPEC 91491 co-author (Europe's first AI-data-readiness standard)
- [Security](https://talonic.com/security)

## Documentation

- [Platform Guide](https://talonic.com/docs): Full product documentation
- [API Reference](https://talonic.com/docs/api): REST endpoint reference
- [Developers](https://talonic.com/developers): Three Modes, billing, MCP, SDKs
- [OpenAPI Spec](https://talonic.com/openapi.json): Note — root path, NOT /docs/openapi.json
- [Full docs](https://talonic.com/llms-full.txt): Concatenated full documentation for deep context