Atom of Thoughts

Structured reasoning for LLMs. Decompose → track confidence → visualize.

---

Setup

Add to your MCP config:

{
  "mcpServers": {
    "atom-of-thoughts": {
      "command": "npx",
      "args": ["-y", "@dioptx/mcp-atom-of-thoughts"]
    }
  }
}

Tip

Works with Claude Code, Cursor, Windsurf, or any MCP client.

How it works

graph LR
    P["P · Premise"]:::premise --> R["R · Reasoning"]:::reasoning
    R --> H["H · Hypothesis"]:::hypothesis
    H --> V["V · Verification"]:::verification
    V --> C["C · Conclusion"]:::conclusion

    classDef premise fill:#6b7280,stroke:#9ca3af,color:#fff,font-weight:bold
    classDef reasoning fill:#3b82f6,stroke:#60a5fa,color:#fff,font-weight:bold
    classDef hypothesis fill:#eab308,stroke:#facc15,color:#000,font-weight:bold
    classDef verification fill:#06b6d4,stroke:#22d3ee,color:#fff,font-weight:bold
    classDef conclusion fill:#22c55e,stroke:#4ade80,color:#fff,font-weight:bold

Loading

Atoms depend on each other, carry confidence scores (0–1), and auto-terminate when a high-confidence conclusion is reached or max depth is hit.

Tools

Tool	Purpose
AoT	Deep reasoning — atoms up to depth 5
AoT-light	Fast reasoning — depth 3, auto-suggests conclusions
atomcommands	Decompose atoms, check termination, adjust depth
export_graph	Get the atom graph as JSON
generate_visualization	Open an interactive D3 graph in your browser
check_approval	Poll for approve/reject decisions from the UI

---

Configuration

Pass flags via args to change behavior:

{
  "args": ["-y", "@dioptx/mcp-atom-of-thoughts", "--mode", "fast", "--no-viz"]
}

Flag	Default	What it does
--mode full|fast|both	both	Which reasoning tools to register
--no-viz	off	Skip visualization and approval tools
--no-approval	off	Skip approval tool only
--max-depth <n>	5 / 3	Override reasoning depth limit
--output-dir <path>	OS temp	Where to write visualization HTML
--downloads-dir <path>	~/Downloads	Where to scan for approval files

Visualization & Approval

generate_visualization creates a self-contained HTML file (D3 bundled inline — works offline) and opens it in your browser. The UI shows:

• Force-directed graph of all atoms and their dependencies
• Color-coded nodes by type with confidence rings
• Sidebar to approve/reject each phase or individual atom
• JSON export that check_approval can poll

No internet connection required — everything is inlined into a single HTML file.

Install Methods

npx (recommended — zero install):

{ "command": "npx", "args": ["-y", "@dioptx/mcp-atom-of-thoughts"] }

npm global:

npm install -g @dioptx/mcp-atom-of-thoughts

{ "command": "mcp-atom-of-thoughts" }

Smithery:

npx -y @smithery/cli install @dioptx/mcp-atom-of-thoughts --client claude

Docker:

docker build -t aot .

{ "command": "docker", "args": ["run", "-i", "--rm", "aot"] }

Development

git clone https://github.com/dioptx/mcp-atom-of-thoughts.git
cd mcp-atom-of-thoughts
npm install
npm test        # 121 tests
npm run build

---

MIT · Based on Atom of Thoughts