DeepL Model Context Protocol (MCP) Server

1.1.0 · active · verified Sun Apr 19

The deepl-mcp-server package provides a Model Context Protocol (MCP) server that integrates DeepL's translation capabilities. It allows other applications, notably Claude Desktop, to leverage DeepL's API for features like text and document translation, rephrasing, automatic language detection, formality control, and glossary support. The current stable version is 1.1.0. This package is primarily designed as a standalone server or CLI tool, rather than a library for programmatic import. Its key differentiator is simplifying DeepL API access for MCP-compatible clients, particularly its direct configuration support for Claude Desktop, enabling AI assistants to perform high-quality translations. Release cadence is infrequent, focusing on stability and integration enhancements. It abstracts the complexities of the DeepL API, offering a streamlined interface via the MCP specification.

Common errors

Warnings

Install

Imports

Quickstart

This quickstart demonstrates how to launch the DeepL MCP server via npx with an API key, then interact with it using curl to call one of its available tools, like fetching source languages.

DEEPL_API_KEY="YOUR_DEEPL_API_KEY" npx deepl-mcp-server &

# Wait a moment for the server to start (it typically runs on port 3000 by default, though not explicitly stated)
# Assuming it runs on default port 3000, as is common for Node.js servers

# Test with a simple request to get available source languages
curl -X POST -H "Content-Type: application/json" \
  -d '{ "tool_code": "get-source-languages", "tool_name": "get-source-languages" }' \
  http://localhost:3000/v1/tools/call

# Example to translate text
# This call structure is based on the MCP specification, which might require a more complex payload
# The actual body for translate-text would involve 'text', 'target_lang', 'source_lang'
# For this quickstart, we'll demonstrate a conceptual call.

# Note: Replace 'YOUR_DEEPL_API_KEY' with an actual key to run successfully.
# The server logs its port on startup; adjust http://localhost:3000 if different.

view raw JSON →