使用说明

datev-mcp

MCP server for DATEV accounting workflows — connect AI agents to German accounting data.

An open-source Model Context Protocol server that enables AI assistants (Claude, ChatGPT, Cursor, etc.) to work with DATEV accounting data: parse EXTF files, look up SKR03/SKR04 chart of accounts, create booking entries, and generate DATEV-compliant exports.

Offline-first. No DATEV account required. No approval process.

Features

EXTF file handling — read, validate, and generate DATEV EXTF booking files
SKR03 & SKR04 — query the standard German charts of accounts (137+ / 89+ accounts)
Booking entries — create and validate DATEV-compliant booking entries
AI suggestions — rule-based account suggestions from invoice text

Quick Start

Claude Desktop

Requires uv installed (provides the uvx command).

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "datev": {
      "command": "uvx",
      "args": ["datev-mcp"]
    }
  }
}

Local Development

git clone https://github.com/BadRix90/datev-mcp
cd datev-mcp
uv venv && uv sync --extra dev
uv run python -m datev_mcp.server

HTTP Mode

uv run python -m datev_mcp.server --transport http
# Server runs on http://localhost:8080

Available MCP Tools

Tool	Description
`datev_buchung_erstellen`	Create a DATEV booking entry with validation
`datev_buchung_vorschlagen`	Suggest accounts from invoice text
`datev_extf_lesen`	Read and parse an EXTF file
`datev_extf_validieren`	Validate an EXTF file for DATEV compliance
`datev_extf_exportieren`	Export booking entries as EXTF file
`datev_kontenrahmen_suchen`	Search SKR03/SKR04 chart of accounts
`datev_kontenrahmen_info`	Get details for a specific account number

MCP Resources

URI	Content
`datev://kontenrahmen/skr03/{nr}`	SKR03 account details
`datev://kontenrahmen/skr04/{nr}`	SKR04 account details

Demo

Web Demo

uv run python demo/server.py
# Open http://localhost:8090

Interactive wizard that demonstrates the AI agent workflow: invoice processing, account lookup, and EXTF export — all powered by real MCP tool calls.

CLI Demo

uv run python demo/mockup.py

Runs 5 scenarios showing how an AI agent would interact with the DATEV MCP server.

Architecture

Phase 1 (current): Offline-first
+-----------+    MCP    +------------+
| AI Agent  | --------> | datev-mcp  |
| (Claude)  |           | Server     |
+-----------+           +-----+------+
                              |
                        +-----v------+
                        | EXTF Parser|
                        | SKR03/04   |
                        | (local)    |
                        +------------+

Phase 2 (planned): DATEV API integration via OAuth 2.0

Configuration

Environment Variable	Description	Default
`DATEV_MCP_LOG_LEVEL`	Log level (DEBUG/INFO/WARNING)	`INFO`
`DATEV_MCP_TRANSPORT`	Transport type (stdio/http)	`stdio`
`DATEV_MCP_PORT`	HTTP port (http transport only)	`8080`

Copy .env.example to .env to configure.

Development

# Run tests
uv run pytest

# Lint & format
uv run ruff check .
uv run ruff format .

Requirements: Python >= 3.12, uv

Tech Stack

FastMCP 3.1.0 — MCP server framework
Pydantic v2 — data validation
Ruff — linting & formatting
pytest — testing

Disclaimer

This project is not affiliated with, endorsed by, or connected to DATEV eG in any way. "DATEV" and "EXTF" are trademarks of DATEV eG. This software implements the publicly documented EXTF file format for interoperability purposes.

The SKR03 and SKR04 chart of accounts data included in this project is based on publicly available information. It is provided for educational and interoperability purposes only. For authoritative account definitions, consult the official DATEV documentation.

This software is provided "as is" without warranty of any kind. It does not constitute tax, legal, or accounting advice. Users are responsible for verifying all generated data before use in production accounting systems.

Author

Kay Dietrich — kaydietrich.com | SyntraQ AI

License

MIT — free to use, modify, and distribute, including commercially.