io.github.rocnubie/deepseekfr-mcp

Official

by rocnubie · JavaScript

DeepSeek FR MCP

Independent French workspace for evaluating DeepSeek-V3-class chat and DeepSeek-R1 reasoning.

Evaluate DeepSeek-V3 and DeepSeek-R1 models for chat and reasoning tasks.

README.md6 KBView on GitHub

DeepSeek FR MCP Server

DeepSeek FR | DeepSeek V3 & R1 workflows in French

A Model Context Protocol server that exposes the canonical DeepSeek FR knowledge surface — models, prompts, and chat workflows, pricing, FAQ, official links — to MCP-compatible AI clients such as Claude Desktop, Cursor, Windsurf, and Continue. Read-only, no API keys, no quota, ~50 ms cold start.

Official website: https://deepseekfr.org

💬 About DeepSeek FR

DeepSeek FR is a French-localized chat interface built around DeepSeek's latest AI models, offering French-speaking users direct access to a capable large language model without needing to navigate an English-first platform. The site provides two model variants — a Pro version aimed at complex reasoning and advanced code tasks, and a Flash version tuned for faster everyday responses — both sharing a one-million-token context window. Users interact through a clean chat interface that supports adjustable reasoning depth, making it practical for everything from quick questions to demanding multi-step analysis. The platform also documents API compatibility with OpenAI and Anthropic conventions, lowering the barrier for developers already familiar with those ecosystems.

Key Features

Two model tiers: Pro for deep reasoning and code-heavy work; Flash for lower-latency, everyday use — both with a 1M token context window.
Three reasoning modes: Non-think (maximum speed), Think High (balanced precision), and Think Max (deepest analysis), selectable per conversation.
Extended context handling: Processes long documents such as contracts, research papers, and technical manuals in a single session without chunking.
API compatibility: Works with OpenAI- and Anthropic-style API clients, so existing tooling can connect with minimal changes.
Benchmark transparency: The site publishes performance data across knowledge, reasoning, code, math, and agent workflow benchmarks so users can evaluate fit before integrating.
Local execution support: Documentation covers running models locally with encoding specifications for teams that prefer on-premise deployments.

Use Cases

Document review: Feeding long contracts, compliance documents, or technical manuals into the chat to extract summaries, flag inconsistencies, or answer specific questions.
Code assistance: Reviewing, debugging, or refactoring code across common languages, with the Pro model handling complex architectural questions.
Research and reasoning: Working through multi-step analytical problems — scientific literature, legal arguments, or financial modeling — where depth matters more than speed.
Content drafting: Generating French-language copy, reports, or structured documents for professional or editorial use.
API prototyping: Testing DeepSeek model responses through a familiar chat interface before wiring them into an application via the compatible API.

Who Is It For

DeepSeek FR is aimed primarily at French-speaking professionals and developers who want a capable AI assistant without switching to an English-language interface. Developers benefit from the API compatibility and local execution options; researchers and analysts value the long context window and adjustable reasoning depth; writers and content teams find the French-native interface reduces friction in day-to-day drafting tasks. The dual-model approach also makes it accessible to non-technical users who need quick answers as well as to technical teams running sustained, resource-intensive workflows — without requiring them to choose a single tool for both.

Tools

`list_models`

Return the canonical list of chat models exposed on the site, with capability notes. (DeepSeek FR)

Input: no parameters. Returns: text/markdown.

`get_pricing`

Return the canonical pricing entry point for DeepSeek FR.

Input: no parameters. Returns: text/markdown.

`get_official_links`

Return the canonical list of official links for DeepSeek FR (website, support, docs when available).

Input: no parameters. Returns: text/markdown.

Resources

site://deepseekfr/models — Supported chat models and capability notes.
site://deepseekfr/pricing — Canonical pricing entry point.
site://deepseekfr/faq — Short FAQ generated from public site metadata.
site://deepseekfr/links — Canonical URLs to share with users.

Prompts

`tell_me_about_deepseekfr`

Summarize what the site is, who it's for, and how it works. — DeepSeek FR

`start_chat_session_deepseekfr`

Open a chat-evaluation session against the site's models, with sensible defaults. — DeepSeek FR

Installation

Install via Smithery

bash

npx -y @smithery/cli install deepseekfr-mcp --client claude

(Replace claude with cursor, windsurf, or continue for those clients.)

Install from source

bash

git clone https://github.com/rocnubie/deepseekfr-mcp.git
cd deepseekfr-mcp
pnpm install

Then add to your MCP client config (claude_desktop_config.json for Claude Desktop, mcp.json for Cursor / Windsurf / Continue):

json

{
  "mcpServers": {
    "deepseekfr-mcp": {
      "command": "node",
      "args": [
        "/absolute/path/to/deepseekfr-mcp/src/index.mjs"
      ]
    }
  }
}

Debug with MCP Inspector

bash

npx @modelcontextprotocol/inspector node src/index.mjs

Official Links

Development

bash

pnpm install
pnpm start                 # run the server over stdio

License

MIT