Access Deepseek Reasoning via API or Ollama
MCP server exposing DeepSeek's reasoning process to Claude and other clients, via the DeepSeek API or a local Ollama server, dual-mode.
1.0.0Add to Favorites
Why it matters
Integrate Deepseek's advanced reasoning and thinking capabilities into your applications. This connector supports both OpenAI API and local Ollama server modes for flexible deployment.
Outcomes
What it gets done
Connect to Deepseek's reasoning engine using OpenAI API.
Integrate with local Ollama server for Deepseek model access.
Capture and utilize structured reasoning processes from Deepseek.
Enable focused reasoning tasks within your existing workflows.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-deepseek-thinker-mcp | bash Capabilities
Tools your agent gets
Performs reasoning using the Deepseek model with structured text response containing the reasoning process.
Overview
deepseek-thinker-mcp server
An MCP server that surfaces DeepSeek's reasoning process to Claude Desktop and other MCP clients, running against either the hosted DeepSeek API or a local Ollama server, through a single get-deepseek-thinker tool. Use it when an AI client needs DeepSeek's own chain-of-thought reasoning as input, rather than only a final answer, whether via the DeepSeek API or a local Ollama deployment.
What it does
An MCP server that captures DeepSeek's reasoning process and passes it to MCP-enabled clients like Claude Desktop, rather than just DeepSeek's final answer. It supports two backends: the hosted DeepSeek API in OpenAI-compatible mode, or a local Ollama server running a DeepSeek model, switchable by environment variable rather than requiring separate installs.
When to use - and when NOT to
Use it when an assistant's own response would benefit from seeing DeepSeek's reasoning trace on a specific prompt - a way to bring a second model's thinking process into context rather than just its conclusion. It is a single-purpose reasoning provider, not a general chat interface to DeepSeek: there is exactly one tool, and it returns reasoning content for one prompt at a time rather than managing a multi-turn conversation. Long or slow reasoning outputs can trigger an MCP request timeout, MCP error -32001, which the project's own FAQ attributes to either a slow DeepSeek API response or unusually long reasoning content, so prompts expected to produce very long chains of thought may need timeout handling on the client side.
Capabilities
get-deepseek-thinker takes one parameter, originPrompt, the user's original prompt, and returns a structured text response containing DeepSeek's reasoning process for that prompt.
How to install
OpenAI API mode:
{
"mcpServers": {
"deepseek-thinker": {
"command": "npx",
"args": [
"-y",
"deepseek-thinker-mcp"
],
"env": {
"API_KEY": "<Your API Key>",
"BASE_URL": "<Your Base URL>"
}
}
}
}
Ollama mode instead sets USE_OLLAMA to true in the env block with no API key needed. Add either configuration to Claude Desktop's claude_desktop_config.json.
Who it's for
Developers building multi-model AI workflows who want DeepSeek's reasoning trace as an input to another assistant's response, whether through the hosted API or a local Ollama deployment. The project is built on TypeScript, the official MCP SDK, and Zod for parameter validation, and is available for local development via a standard npm install and build. The project is licensed under MIT.
Source README
Deepseek Thinker MCP Server
A MCP (Model Context Protocol) provider Deepseek reasoning content to MCP-enabled AI Clients, like Claude Desktop. Supports access to Deepseek's thought processes from the Deepseek API service or from a local Ollama server.
Core Features
🤖 Dual Mode Support
- OpenAI API mode support
- Ollama local mode support
🎯 Focused Reasoning
- Captures Deepseek's thinking process
- Provides reasoning output
Available Tools
get-deepseek-thinker
- Description: Perform reasoning using the Deepseek model
- Input Parameters:
originPrompt(string): User's original prompt
- Returns: Structured text response containing the reasoning process
Environment Configuration
OpenAI API Mode
Set the following environment variables:
API_KEY=<Your OpenAI API Key>
BASE_URL=<API Base URL>
Ollama Mode
Set the following environment variable:
USE_OLLAMA=true
Usage
Integration with AI Client, like Claude Desktop
Add the following configuration to your claude_desktop_config.json:
{
"mcpServers": {
"deepseek-thinker": {
"command": "npx",
"args": [
"-y",
"deepseek-thinker-mcp"
],
"env": {
"API_KEY": "<Your API Key>",
"BASE_URL": "<Your Base URL>"
}
}
}
}
Using Ollama Mode
{
"mcpServers": {
"deepseek-thinker": {
"command": "npx",
"args": [
"-y",
"deepseek-thinker-mcp"
],
"env": {
"USE_OLLAMA": "true"
}
}
}
}
Local Server Configuration
{
"mcpServers": {
"deepseek-thinker": {
"command": "node",
"args": [
"/your-path/deepseek-thinker-mcp/build/index.js"
],
"env": {
"API_KEY": "<Your API Key>",
"BASE_URL": "<Your Base URL>"
}
}
}
}
Development Setup
# Install dependencies
npm install
# Build project
npm run build
# Run service
node build/index.js
FAQ
Response like this: "MCP error -32001: Request timed out"
This error occurs when the Deepseek API response is too slow or when the reasoning content output is too long, causing the MCP server to timeout.
Tech Stack
- TypeScript
- @modelcontextprotocol/sdk
- OpenAI API
- Ollama
- Zod (parameter validation)
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.