Happy slice of mcp pizzaMCP.pizza

Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP

MCP.Pizza Chef: newideas99

Ask it a tough question and it runs two models in sequence through OpenRouter: DeepSeek R1 works out the reasoning, then Claude writes the final answer using that thinking. Two things to know. The models are fixed in the code, so the settings the README calls optional model configuration are never actually read, and you are pinned to Claude 3.5 Sonnet from late 2024. By default it also reads your most recent Cline conversation off disk and sends it to both models. No code changes since February 2025.

Unmaintained · No commits in 19 months.
Coding
Writing

Use This MCP server To

Think through a tricky design decision before I commit to it See the reasoning steps alongside the final answer Ask a hard question that keeps tripping up my usual assistant Carry on from where my last coding conversation left off

README

Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP

smithery badge

A Model Context Protocol (MCP) server that combines DeepSeek R1's reasoning capabilities with Claude 3.5 Sonnet's response generation through OpenRouter. This implementation uses a two-stage process where DeepSeek provides structured reasoning which is then incorporated into Claude's response generation.

Features

  • Two-Stage Processing:

    • Uses DeepSeek R1 for initial reasoning (50k character context)
    • Uses Claude 3.5 Sonnet for final response (600k character context)
    • Both models accessed through OpenRouter's unified API
    • Injects DeepSeek's reasoning tokens into Claude's context

  • Smart Conversation Management:

    • Detects active conversations using file modification times
    • Handles multiple concurrent conversations
    • Filters out ended conversations automatically
    • Supports context clearing when needed

  • Optimized Parameters:

    • Model-specific context limits:
      • DeepSeek: 50,000 characters for focused reasoning
      • Claude: 600,000 characters for comprehensive responses
    • Recommended settings:
      • temperature: 0.7 for balanced creativity
      • top_p: 1.0 for full probability distribution
      • repetition_penalty: 1.0 to prevent repetition

Installation

Installing via Smithery

To install DeepSeek Thinking with Claude 3.5 Sonnet for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @newideas99/Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP --client claude

Manual Installation

  1. Clone the repository:
git clone https://github.com/yourusername/Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP.git
cd Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP
  1. Install dependencies:
npm install

Deepseek-Thinking-Claude-3.5-Sonnet-CLINE-MCP FAQ

Do I need a key?
Yes, an OpenRouter key on an account with a balance. Every question bills you twice, once for the reasoning pass and once for the answer, so costs add up faster than a normal chat.
Can I choose which models it uses?
No. The README lists model names as optional settings, but both are written directly into the code and those settings are never read. You get DeepSeek R1 and Claude 3.5 Sonnet.
Does it send my conversation history anywhere?
Yes, by default. It reads your most recent Cline conversation off disk and includes up to 50,000 characters in the reasoning request and up to 600,000 in the answer request. Set includeHistory to false to stop that.
Which apps does it work in?
It is built for Cline in VS Code. The history feature only works there, though the basic question tool runs in Claude Desktop or any assistant that can start a local command.
Can I use this to reason through a hard problem?
Yes — that is the whole point. You ask once, the reasoning model thinks it through, and the second model writes the reply with that thinking in front of it.
Why does a reply take so long?
Two models run one after the other. The README warns replies can take up to a minute, so it hands back a task number and you check on it.
How hard is setup?
Developer-level. You clone the project, install its packages and build it before you can paste the configuration block into your editor.
Is it still maintained?
There is no sign of it. The last code change was February 2025, around eighteen months ago, and it is still pinned to a model released in October 2024.