Think MCP

What is Think MCP about?

Think MCP delivers an MCP server that equips AI agents with a “think” tool for explicit thought logging, facilitating clearer reasoning, backtracking, and policy compliance during complex workflows.

How to use Think MCP?

1. Add the server configuration to your agent’s MCP settings, e.g.:

"mcpServers": {
  "think-mcp": {
    "command": "uvx",
    "args": ["think-mcp"],
    "enabled": true
  }
}

2. Invoke the think tool by sending a payload with a thought string. In advanced mode, additional tools (criticize, plan, search) become available via the --advanced flag and require a TAVILY API key.
3. The server appends each thought to the agent’s log, which can be consulted in subsequent steps.

Key features of Think MCP

• Implements the “think” tool as described by Anthropic, supporting explicit thought recording.
• Minimal, standards‑based MCP implementation using the mcp Python package.
• Advanced mode adds supplementary tools (criticize, plan, search).
• Simple configuration via command‑line invocation (uvx think-mcp or uvx think-mcp --advanced).

Use cases of Think MCP

• Tool output analysis – pause after a tool call to reflect on results before proceeding.
• Policy‑heavy environments – log compliance checks as thoughts to ensure guidelines are followed.
• Sequential decision making – maintain a reasoning trail across multi‑step processes.
• Research and benchmarking – evaluate the impact of explicit reasoning on model performance.

FAQ from the Think MCP

Q: Do I need a specific model to use the think tool? A: No, the tool works with any model that can interact via MCP; it adds reasoning capability externally.

Q: How does advanced mode differ from the basic mode? A: Advanced mode enables additional tools (criticize, plan, search). The search tool requires a TAVILY API key set in the environment.

Q: Can I run Think MCP locally? A: Yes, install the think-mcp package and start the server with uvx think-mcp (or uvx think-mcp --advanced).

Q: What format should the thought input take? A: A simple string describing the agent’s current consideration or decision point.

Think MCP Tool

Think MCP is an implementation of an MCP (Model Context Protocol) server that provides a "think" tool for structured reasoning in agentic AI workflows. This project is inspired by the Anthropic engineering article: The "think" tool: Enabling Claude to stop and think in complex tool use situations.

According to the referenced article, adding the think tool can lead to improved evaluation metrics by enabling reasoning capabilities even in models that do not natively possess advanced reasoning skills.

What is the "think" tool?

The "think" tool allows an AI agent to pause and record an explicit thought during complex reasoning or multi-step tool use. It does not change the environment or database, but appends the thought to the log, helping the agent process information, backtrack, or comply with detailed policies.

This approach is especially useful for:

• Tool output analysis (processing results of previous tool calls)
• Policy-heavy environments (verifying compliance with guidelines)
• Sequential decision making (where each step builds on previous ones)

Features

• Implements the "think" tool as described in Anthropic's research
• Minimal, standards-based MCP server using mcp[cli]
• Ready for integration with Claude or other agentic LLMs

Usage

MCP server configuration

Add this MCP server to your facorite agent.

"mcpServers": {
    "think-mcp": {
        "command": "uvx",
        "args": ["think-mcp"],
        "enabled": true
    }
}

Tool definition

The "think" tool is defined as:

• Input: thought (string) — A thought to think about.
• Behavior: Appends the thought to the log for structured reasoning.

Advanced mode

Adds aditional tools for your agent:

• criticize
• plan
• search

"mcpServers": {
    "think-mcp": {
        "command": "uvx",
        "args": ["think-mcp", "--advanced"],
        "enabled": true,
        "env": {
            "TAVILY_API_KEY": ... YOUR TAVILY API KEY HERE ...
        }
    }
}

Reference

• Based on: Anthropic Engineering Blog — The "think" tool

License

MIT License — see LICENSE