What’s New (2025-04):
- Advanced visualization: clustering (k-means/degree), centrality overlays, edge bundling, and agentic overlays for tasks and priorities
- Agentic cache & prefetch: LRU+TTL caches for embeddings, summaries, analytics, and proactive agent cache warming
- Enhanced analytics: real-time, multi-branch, and focusNode support; agent-optimized metadata
- Upgraded documentation and onboarding for agents and users
Features
- 🌳 Branch Management: Create, focus, and navigate multiple lines of thought
- 🔗 Cross-References: Link related thoughts across branches (typed, scored)
- 💡 AI Insights: Automatic insight and summary generation
- 🧠 Semantic Search: Find related thoughts using embeddings
- 📊 Advanced Visualization:
- Node clustering (k-means/degree)
- Centrality overlays (closeness, betweenness)
- Edge bundling
- Task overlays (status, priority, next-action)
- Agentic overlays and metadata for all nodes/edges
- FocusNode and multi-branch visualization
- ⚡ Agentic Cache & Prefetch:
- LRU+TTL caches for embeddings, summaries, analytics
- Proactive cache warming for agent workflows
- 🗂️ Persistent Storage: Queryable, extensible, and never lose a thought
- 🔄 Real-Time & Multi-Branch: Visualize and analyze multiple branches and nodes in real-time
- 🛠️ Production-Grade: Robust error handling, performance optimizations, and agent/human-friendly APIs
🛠️ Technology Stack
- Node.js (18+)
- TypeScript (4.x)
- @dagrejs/graphlib: Graph structure, algorithms, and analytics
- ml-kmeans: Clustering for visualization
- lru-cache: LRU+TTL caching for embeddings, summaries, analytics
- @xenova/transformers: Embedding and summarization pipelines
- @modelcontextprotocol/sdk: MCP protocol and agent integration
- chalk: CLI output styling
- Mermaid: Gantt/roadmap visualization
- pnpm: Fast dependency management
Roadmap (Gantt)
Table of Contents
Why Branch-Thinking?
- Agentic by Design: Built for both human and AI workflows—every command is agent-friendly.
- True Branching: Organize, cross-link, and reason over ideas, code, and tasks in parallel.
- AI-Native: Semantic search, auto-summarization, and insight generation out-of-the-box.
- Persistent & Visual: Never lose a thought—everything is queryable, visualizable, and extensible.
Summary
Branch-Thinking MCP Tool is an advanced agentic platform for managing, visualizing, and reasoning over branching thoughts, tasks, code, and knowledge. It empowers both AI agents and humans to organize complex projects, cross-link ideas, and automate insight generation using a powerful branch-based paradigm. With semantic search, visualization, and persistent task/code management, it is designed for next-generation collaborative and autonomous workflows.
Branch-Thinking MCP Tool is an advanced agentic platform for managing, visualizing, and reasoning over branching thoughts, tasks, code, and knowledge. It empowers both AI agents and humans to organize complex projects, cross-link ideas, and automate insight generation using a powerful branch-based paradigm. With semantic search, visualization, and persistent task/code management, it is designed for next-generation collaborative and autonomous workflows.
Architecture & Flow
Quick Start
Get up and running in seconds:
pnpm install # Recommended for speed (or npm install)
pnpm build
node dist/index.js --help # See available commands
Getting Started
1. Clone & Install
git clone https://github.com/your-org/branch-thinking-mcp.git
cd branch-thinking-mcp
pnpm install # Or npm install
pnpm build # Or npm run build
For Claude Desktop integration, add to your claude_desktop_config.json:
"branch-thinking": {
"command": "node",
"args": [
"/your-custom-mcp-dir-here/branch-thinking/dist/index.js"
]
}
3. Run
Real-World Usage Recipes
1. Knowledge Capture & Linking
# Batch capture meeting notes
add-thought dev "Discussed semantic search improvements" note
add-thought dev "Agreed to refactor API" decision
# Link related thoughts
link-thoughts t1 t2 supports "API refactor supports search improvements"
# Extract and manage tasks from a research branch
extract-tasks research
list-tasks research open
update-task-status task-1 in_progress
3. Visualization for Insight
# Generate and interpret a knowledge graph
visualize dev
# Review AI-generated summary
summarize-branch dev
🧑💻 Live Example: Agentic Workflow
# 1. Create a new branch for your project or idea
create-branch "AI Research"
# 2. Add thoughts and observations
add-thought [branchId] "Explore semantic search for agent workflows" analysis
add-thought [branchId] "Test cross-linking and summarization" observation
# 3. Link related thoughts
link-thoughts [thoughtId1] [thoughtId2] supports "Thought 2 validates Thought 1"
# 4. See your knowledge graph
visualize [branchId]
# 5. Extract tasks and get AI review
extract-tasks [branchId]
review-branch [branchId]
Replace [branchId] and [thoughtIdX] with actual IDs from list and history.
Command Reference
Branch Management
| Command | Description |
|---|
list | Show all branches with status |
focus [branchId] | Switch focus to a branch |
history [branchId?] | Show thought history |
summarize-branch [branchId?] | AI summary of branch |
review-branch [branchId?] | AI review of branch |
visualize [branchId?] | Visual graph of connections |
Thought & Insight Management
| Command | Description |
|---|
insights [branchId?] | Get AI-generated insights |
crossrefs [branchId?] | Show cross-references |
hub-thoughts [branchId?] | List hub thoughts |
semantic-search [query] | Find similar thoughts |
link-thoughts [from] [to] [type] [reason?] | Link two thoughts |
add-snippet [content] [tags] | Save a code snippet |
snippet-search [query] | Search code snippets |
doc-thought [thoughtId] | Document a thought |
Task Management
| Command | Description |
|---|
extract-tasks [branchId?] | Extract actionable items |
list-tasks [branchId] [status] [assignee] [due] | List/filter tasks |
update-task-status [taskId] [status] | Update a task’s status |
summarize-tasks [branchId] | Summarize tasks |
AI & Knowledge
| Command | Description |
|---|
ask [question] | AI answer from knowledge base |
Best Practices
- Always start with
create-branch to ensure clean context. - Use
list and focus to navigate between projects or lines of thought. - Leverage
summarize-branch and insights after adding several thoughts to get AI-generated context. - Use
link-thoughts to explicitly connect ideas, tasks, or code for richer semantic graphs. - After code changes, always run
pnpm lint and pnpm build to catch errors early. - Decompose complex goals into sequences of thought/task/insight commands.
- Iterate and adapt: Use feedback from summaries, reviews, and visualizations to refine next actions.
- Explicitly specify parameters (branchId, status, assignee, etc.) for precise results.
- Use cross-references and multi-hop links to foster creativity and bridge ideas.
- Prompt agents (Claude, GPT-4, etc.) to "think step by step" or "use chain of thought" for best results.
Security
- All persistent data is stored locally (default: project directory or
MCP_STORAGE_PATH) - No external API calls unless configured
- Agents/users are responsible for privacy of stored thoughts and tasks
- To report security issues, please open an issue or email the maintainer.
Troubleshooting and FAQ
Q: The tool isn't responding!
A: Check the MCP server logs and ensure configuration is correct.
Q: How do I reset storage?
A: Delete or move the persistent storage directory (see config).
Q: How do I add a new command?
A: Extend handleCommand in src/index.ts and document it in the README.
Accessibility and Internationalization
- All badges/images have descriptive alt text.
- English is the default language; contributions for translations are welcome.
- Please open a PR or issue if you want to help localize this tool.
Contributing
Contributions, issues, and feature requests are welcome! Please open a PR or issue on GitHub.
- Fork this repo
- Create a new branch (
git checkout -b feature/your-feature) - Commit your changes
- Push to the branch
- Open a Pull Request
Credits
- Concept & Testing: @ssdeanx
- Core Code Generation: Claude, GPT-4, and Cascade
- Implementation, Fixes, and Documentation: @ssdeanx
License
MIT