# dokumendiregister MCP — Estonian public-authority documents for AI assistants

[![smithery badge](https://smithery.ai/badge/dokumendiregister/dokumendiregister)](https://smithery.ai/servers/dokumendiregister/dokumendiregister)

Connect any MCP-compatible AI assistant (Claude, Cursor, ChatGPT, and more) to the
**document registers of Estonian public authorities** — ministries, agencies,
inspectorates and municipalities. Search every register at once, read a document's
metadata and extracted file text, or pull every document naming a given company — in
plain language, **free, no API key, read-only**.

Powered by **[dokumendiregister.ee](https://dokumendiregister.ee)**, a free, ad-free
aggregator that indexes those registers together and links every document to the
companies named in it (via the Estonian Business Register / [nimistu.ee](https://nimistu.ee)).

- **Endpoint (Streamable HTTP):** `https://dokumendiregister.ee/api/dokreg-mcp/mcp`
- **Free** — no API key, no signup, read-only
- **Cross-register** — one search across all indexed authorities
- **Full-text** — searches the extracted text of attached files, not just metadata
- **Company-linked** — every document ties to the companies it names

## Why this over browsing a register by hand

Each Estonian authority runs its own document register, and there is no shared search
across them. That makes the documents public in principle but hard to actually find:

- **One search, all authorities** — a single query spans every indexed register instead
  of one authority's search form at a time.
- **Company ↔ document join** — `documents_by_company` returns every public-authority
  document naming a company by its 8-digit registry code, across all registers, with the
  company's role. No other Estonian tool exposes this.
- **File text** — PDFs, ASiC-E/BDoc containers and e-mail attachments are text-extracted
  and searchable, not just linked.
- **ChatGPT-compatible** `search` + `fetch`, so the same server works across clients.

## Quick start

### Claude Code

```bash
claude mcp add --transport http dokumendiregister https://dokumendiregister.ee/api/dokreg-mcp/mcp
```

### Cursor

Add to `~/.cursor/mcp.json` (global) or `.cursor/mcp.json` (project):

```json
{
  "mcpServers": {
    "dokumendiregister": {
      "url": "https://dokumendiregister.ee/api/dokreg-mcp/mcp"
    }
  }
}
```

### Claude Desktop and other clients

```json
{
  "mcpServers": {
    "dokumendiregister": {
      "url": "https://dokumendiregister.ee/api/dokreg-mcp/mcp"
    }
  }
}
```

Stdio-only clients can bridge via [`mcp-remote`](https://www.npmjs.com/package/mcp-remote):

```json
{
  "mcpServers": {
    "dokumendiregister": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://dokumendiregister.ee/api/dokreg-mcp/mcp"]
    }
  }
}
```

**Transports:** Streamable HTTP at `https://dokumendiregister.ee/api/dokreg-mcp/mcp`
(recommended) · SSE at `https://dokumendiregister.ee/api/dokreg-mcp/sse`.

## What you can ask

You talk to it in natural language — you don't need to know reference numbers or tool
details.

- Which public-authority documents mention registry code 16075394?
- Find documents about `rahapesu` (money laundering) and show me the newest
- Show me the full text and attachments of that document
- What has the Financial Intelligence Unit (Rahapesu Andmebüroo) registered this month?
- Incoming letters to the Registrite ja Infosüsteemide Keskus since 2026-01-01
- Which authorities are indexed, and how many documents does each have?

## Tools

Six tools over the full dokumendiregister.ee corpus:

| Tool | What it does |
| --- | --- |
| **search** | Full-text document search across all authorities; returns matches with a stable id + dokumendiregister.ee URL (ChatGPT-compatible) |
| **fetch** | A document's metadata, file list and extracted text by id (ChatGPT-compatible) |
| **search_documents** | Filtered search: authority, document type, access restriction, registration-date range |
| **documents_by_company** | Every document naming a company, by registry code, across all registers — unique to dokumendiregister.ee |
| **list_authorities** | All indexed authorities, with document counts |
| **recent_documents** | The newest registered documents |

## Data sources and attribution

Documents come from the **authorities' own public document registers**, reused as open
data. Access-restricted (`AK`) documents expose metadata only — never file contents.
Company links come from [nimistu.ee](https://nimistu.ee), which aggregates the Estonian
Business Register and related open data.

- Website — <https://dokumendiregister.ee>
- MCP connection page — <https://dokumendiregister.ee/mcp>
- Company data — <https://nimistu.ee>

## License

The configuration examples and documentation in this repository are released under the
[MIT License](LICENSE). The dokumendiregister.ee service and the underlying registers
have their own terms.