Summary
Implements Anthropic's think tool as an MCP server, giving Claude a structured workspace for complex reasoning tasks. Exposes a single think operation that lets Claude break down problems step-by-step, plus utilities to view thought history, get thinking statistics, and clear the slate. Works with Claude Desktop, Windsurf, and Cursor. Grab this when you're tackling multi-step workflows or policy-heavy tasks where Claude needs to show its work and maintain reasoning chains across long tool sequences.
Give your AI agent a complete email layer—sending, inbound inboxes, and sandbox testing.
Get 4K emails/month free →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Capacitor makes coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
MCP Think Tool Server
A Model Context Protocol (MCP) server implementing the "think" tool for improving Claude's complex reasoning capabilities.
Overview
This MCP server implements the "think" tool as described in Anthropic's blog post, which provides Claude with a dedicated space for structured thinking during complex problem-solving tasks. The think tool has been shown to significantly improve performance in complex tasks requiring policy adherence and reasoning in long chains of tool calls.
Features
- Structured Thinking Space: Provides Claude with a dedicated place to break down complex problems
- Thought History: Maintains a log of all thoughts with timestamps for reference
- Statistics and Analysis: Offers metadata about thinking patterns
- Clean Slate Option: Allows clearing thought history when starting fresh
Installation
Install from PyPI:
pip install mcp-think-tool
Configuration
Windsurf
To use this tool with Claude in Windsurf, add the following configuration to your MCP config file:
"think": {
"command": "/home/xxx/.local/bin/mcp-think-tool",
"args": [],
"type": "stdio",
"pollingInterval": 30000,
"startupTimeout": 30000,
"restartOnFailure": true
}
The command field should point to the directory where you installed the python package using pip.
Docker
You can install this MCP server with only the Dockerfile
First download the Dockerfile, navigate to its directory, and build the Docker image
docker build -t mcp-think-tool .
Then add the following configuration your MCP config file
"think": {
"command": "docker",
"args": ["run", "--rm", "-i", "mcp-think-tool"]
}
This was tested and working with Claude Desktop and Cursor
Featured
Give your AI agent a complete email layer—sending, inbound inboxes, and sandbox testing.
Get 4K emails/month free →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →Capacitor makes coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →UpdatedMar 9, 2026