![\"GoT](\"docs/assets/hero-complex-graph.png\")
\r\n\r\n \r\n
\r\n Graph of Thoughts MCP Server — Bounded, Auditable Reasoning for AI Agents\r\n
\r\n\r\n\r\n ![\"npm](\"https://img.shields.io/npm/v/@abderraouf-yt/got-mcp?style=flat-square&color=00e5ff&label=npm\")
![\"GitHub](\"https://img.shields.io/github/stars/Abderraouf-yt/got-mcp?style=flat-square&color=ff00ff&label=stars\")
![\"Node.js\"](\"https://img.shields.io/badge/node-%3E%3D20-00e5ff?style=flat-square\"){kind=icon}
![\"MCP](\"https://img.shields.io/badge/MCP-1.26+-ff00ff?style=flat-square\"){kind=icon}
![\"Tools\"](\"https://img.shields.io/badge/tools-16-00ff88?style=flat-square\"){kind=icon}
![\"License\"](\"https://img.shields.io/badge/license-MIT-yellow?style=flat-square\"){kind=icon}
![\"GoT](\"https://img.shields.io/badge/GoT-Besta%20et%20al.%202023-ff6b6b?style=flat-square\"){kind=icon}
\r\n ⚡ Quick Start •\r\n 🛠 Tools •\r\n 🧬 How It Thinks •\r\n 🔒 Governance •\r\n 📊 Visualizer\r\n
\r\n\r\n---\r\n\r\n## What is this?\r\n\r\n> **Chain of Thought** reasons in a line. **Tree of Thought** can branch. **Graph of Thoughts** can branch, merge, contradict, prune dead ends, and converge on the winning path.\r\n\r\nThought Graph is a [Model Context Protocol (MCP)](https://modelcontextprotocol.io) server that gives AI agents **non-linear reasoning** — a directed acyclic graph where thoughts can branch into alternatives, get contradicted by evidence, and converge through weighted aggregation.\r\n\r\n```\r\nTraditional AI: A → B → C → D (linear, no backtracking)\r\nTree of Thought: A → B₁, B₂ (branching, no merging) \r\nGraph of Thoughts: A → B₁ ←contradiction→ C₁, B₂ →aggregate→ D ← THIS\r\n```\r\n\r\n| 🔗 Chain of Thought Linear • No backtrack | \r\n🌲 Tree of Thoughts Branching • No merge | \r\n🧠 Graph of Thoughts Branch • Merge • Prune • Converge | \r\n
| Tool | What it does | Category |
|---|---|---|
propose_thought | Add a reasoning node with optional parent + relation | 🧠 Core |
evaluate_thought | Score (0.0–1.0) + critique. Omit score → autonomous LLM audit | 🧠 Core |
get_thought_graph | Retrieve the full DAG state | 🧠 Core |
reset_graph | Clear graph for new session | 🧠 Core |
aggregate_thoughts | Merge 2+ nodes → weighted synthesis Σ(score×weight)/Σ(weights) | ⚡ GoT |
prune_branch | Hard (reject+zero) or Soft (score decay) pruning | ⚡ GoT |
find_winning_path | Greedy DFS or beam search (k-best paths) with threshold gating | ⚡ GoT |
get_graph_metrics | Node count, max depth, prune ratio, avg score, status breakdown | 📊 Ops |
export_snapshot | Full graph serialization for replay/recovery | 🔁 Replay |
restore_snapshot | Restore from previously exported snapshot | 🔁 Replay |
reflect_and_refine | Self-reflection: 4-axis confidence + auto-critique + branch | 🔬 v4.0 |
context_set | Write key-value to shared context store with provenance | 📦 v4.0 |
context_get | Read value + source from shared context store | 📦 v4.0 |
context_list | List all context store entries and their sources | 📦 v4.0 |
export_reasoning_trace | Export winning path as Long CoT trace (DeepSeek-R1/o3 format) | 📤 v4.0 |
run_controller_loop | Autonomous GoT orchestrator: seed → evaluate → branch → reflect → prune → converge | 🔄 v4.0 |
\r\n Evolution of sequential thinking — because the best ideas don't come in a straight line.\r\n
\r\n","readmeFilename":"README.md"}