hubinoretros/deep-thinker

Overview

An MCP server that implements a directed acyclic graph (DAG) thought graph for structured reasoning. It applies 10 reasoning strategies, metacognition, self-critique, knowledge integration, and pruning to refine outputs.

Best for

Best for
Developers building advanced reasoning agents within the MCP ecosystem

Use cases

• Building multi-step reasoning chains for LLM agents
• Implementing self-correcting thought processes in MCP hosts
• Combining multiple reasoning strategies for complex queries

How to use

Install

npx deep-thinker

Tools exposed

• Self-critique

Tested with

Claude Desktop

Notes

An MCP server that implements a directed acyclic graph (DAG) thought graph for structured reasoning. It applies 10 reasoning strategies, metacognition, self-critique, knowledge integration, and pruning to refine outputs.

7 stars on GitHub. Last updated 2026-05-23. Licensed MIT.

Use cases

• Building multi-step reasoning chains for LLM agents
• Implementing self-correcting thought processes in MCP hosts
• Combining multiple reasoning strategies for complex queries

Pros

• Structured DAG-based reasoning improves coherence
• Includes self-critique and pruning to reduce errors
• Supports multiple reasoning strategies in one tool

Cons

• Low community adoption with only 7 stars
• Requires an MCP-compatible host to function
• Setup and configuration may be non-trivial

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.