# DocQuery

> DocQuery is a universal document-to-API platform. Upload spreadsheets, PDFs,
> contracts, invoices, or any document; an AI engine extracts structured
> relational tables; you link them with foreign keys; and DocQuery serves them
> as REST, GraphQL, and MCP APIs.

The product turns scattered files (Excel, CSV, PDF, Word, Markdown, JSON,
images, scanned forms) **and live Airtable bases** into a queryable
relational backend with three parallel API surfaces:

- **REST** with OpenAPI 3.1: per-table CRUD, filtering, pagination, and
  relation expansion via `include=`.
- **GraphQL** (Absinthe-based): workspace-aware schema with nested relations
  and typed mutations.
- **MCP** (Model Context Protocol): JSON-RPC 2.0 endpoint exposing tools,
  resources, and prompts to any MCP-aware AI agent (Claude, Cursor, etc.).

## Input sources

- File upload: CSV, TSV, JSON, Markdown, plain text. PDF, Excel, Word
  on the roadmap.
- Paste content: same formats, pasted directly into a textarea.
- **Airtable**: connect a base by Personal Access Token (PAT). DocQuery
  walks Airtable's pagination, extracts every record, infers column
  types (number / boolean / date / string), and stores the Airtable
  record id as the primary key. Re-pull on demand to refresh.

Each user gets a free workspace on signup. Plans:
- Free: 5 documents, 10,000 rows, basic extraction.
- Starter ($29/mo): 50 documents, 250,000 rows, live sync, full APIs.
- Pro ($79/mo): unlimited (fair use), advanced OCR + LLM, RLS, priority.
- Enterprise: custom SLAs, on-prem, dedicated parsing models.

## Marketing pages

- [Landing](/): product overview, hero, how-it-works, triple-API, use cases.
- [Features](/features): detailed feature list with examples.
- [How it works](/how-it-works): six-step walkthrough + worked example.
- [Pricing](/pricing): plans, FAQ.
- [About](/about): company beliefs and contact.
- [Privacy Policy](/privacy)
- [Terms of Service](/terms)

## Product surfaces (require account)

- /dashboard — workspace overview
- /documents — upload + list source documents
- /tables — extracted relational tables
- /relations — foreign-key canvas
- /api-docs — REST + GraphQL + MCP try-it console
- /chat — AI chat backed by the workspace's tables
- /billing — plan + Stripe customer portal
- /workspaces — multi-workspace settings
- /account — email + password settings

## Public APIs (require workspace API key as Bearer token)

- `GET /api/v1/tables` — list tables in the workspace.
- `GET /api/v1/tables/:slug/rows?include=col` — filter + expand relations.
- `POST /api/v1/graphql` — Absinthe GraphQL endpoint.
- `POST /api/v1/mcp` — Model Context Protocol JSON-RPC.
- `GET /api/v1/openapi.json` — OpenAPI 3.1 spec for the workspace.
- `GET /api/v1/docs` — Redoc-rendered HTML viewer (auth required).

## For AI agents

Point your MCP-capable client at `/api/v1/mcp` with header
`Authorization: Bearer <workspace_api_key>`. The server implements
`initialize`, `tools/list`, `tools/call`, `resources/list`, `resources/read`,
`prompts/list`, `prompts/get`. Tools: `list_tables`, `list_relations`,
`query_table`. Resources: `docquery://schema` and `docquery://tables/<slug>`.

## Source &amp; status

- Built on Phoenix LiveView + SQLite, source-deployed via systemd + Cloudflare Tunnel.
- AI extraction powered by a state-of-the-art LLM with structured outputs.
- Stripe-backed subscriptions; webhook signature verification built in.