Which integrations are available for this server?

Supports exporting thought sessions in [Markdown](/mcp/servers/integrations/markdown) format for human-readable documentation of reasoning processes.

🧠 Think MCP

MCP Server for structured sequential thinking. Helps LLMs break down complex problems into manageable steps with branching, revisions, and self-critique.

Features

Sequential Thinking — Step-by-step reasoning with confidence tracking
Branching & Revisions — Explore alternatives, fix mistakes
Dead Ends Tracking — Remember rejected paths to avoid circular thinking
Proactive Coach — Nudges for better thinking (low confidence, missing critique)
Fuzzy Recall — Search through thought history with Fuse.js

Quick Start

Using npx (recommended)

npx @gofman3/think-mcp

Install globally

npm install -g @gofman3/think-mcp think-mcp

MCP Configuration

Add to your MCP config (mcp.json):

{ "mcpServers": { "think": { "command": "npx", "args": ["@gofman3/think-mcp"] } } }

Or if installed globally:

{ "mcpServers": { "think": { "command": "think-mcp" } } }

Tools

`sequentialthinking`

Primary tool for step-by-step reasoning.

{ thought: string, // Your current reasoning step thoughtNumber: number, // Sequential: 1, 2, 3... totalThoughts: number, // Estimate (auto-adjusts) confidence?: number, // 1-10 nextThoughtNeeded: boolean, isRevision?: boolean, revisesThought?: number, branchFromThought?: number, branchId?: string, subSteps?: string[], // Micro-plan (max 5) alternatives?: string[] // Quick comparison (max 5) }

`extend_thought`

Deep-dive into a specific thought without advancing.

{ targetThoughtNumber: number, extensionType: 'critique' | 'elaboration' | 'optimization' | 'polish' | ..., content: string, impactOnFinalResult: 'low' | 'medium' | 'high' | 'blocker' }

`consolidate_and_verify`

Final validation before answering.

{ winningPath: number[], // e.g., [1, 2, 5, 8] summary: string, verdict: 'ready' | 'needs_more_work' }

`recall_thought`

Fuzzy search through thought history.

{ query: string, scope?: 'current' | 'all', searchIn?: 'thoughts' | 'extensions' | 'alternatives' | 'all', limit?: number, threshold?: number // 0-1, lower = stricter }

`reset_session`

Clear all thoughts and start fresh.

`export_session`

Export session as Markdown or JSON.

License

MIT

This server cannot be installed

-

security - not tested

A

license - permissive license

-

quality - not tested

How are these scores calculated?

Resources

GitHub Repository

Need Help?

Report Issue

Related Servers

Think MCP