Production-grade MCP servers for the agentic future Semantic search • Code analysis • AI orchestration • Verifiable provenance
Website: mcptoolshop.com • Philosophy: Syntropy above all else
┌─────────────────────────────────────────────────────────────────┐
│ 🔐 CID Registry Spec │
│ Content-Addressed Tool Registry with Crypto Verify │
│ (24 test vectors • Ed25519 • dag-cbor CIDs) │
└──────────────┬──────────────────────────────────┬──────────────┘
│ │
┌──────────────▼──────────────────────────────────▼──────────────┐
│ 🧭 Tool Compass │
│ Semantic MCP Tool Discovery Engine │
│ (115+ tools • 6 backends • HNSW search) │
└──────────────┬──────────────────────────────────┬──────────────┘
│ │
┌──────────▼──────────┐ ┌──────────▼──────────┐
│ 📁 File Compass │ │ 🧠 Dev-Brain │
│ Semantic File │ │ Code Analysis │
│ Search Engine │ │ & Health Scoring │
└─────────────────────┘ └─────────────────────┘
│ │
┌──────────▼──────────┐ ┌──────────▼──────────┐
│ 🎨 Comfy-Headless │ │ 🔬 Backpropagate │
│ AI Image/Video │ │ LLM Fine-Tuning │
│ Generation │ │ Framework │
└─────────────────────┘ └─────────────────────┘
│
┌──────────▼──────────┐
│ 🎛️ Integradio │
│ Vector-Embedded │
│ UI Components │
└─────────────────────┘
Content-Addressed Tool Registry
- Cryptographic tool verification via CIDs
- Ed25519 signatures with verifiable provenance
- 24 conformance test vectors
- Deterministic 9-step verification pipeline
- Status: ✅ v0.1.1 Released • Docs: Verifier Rulebook
Formal Provenance Specification
- Language-neutral provenance standard
- Cryptographic attestation chains
- Status: ✅ Production • Stack: Spec + TypeScript reference
Semantic MCP Tool Discovery
- Find tools by intent, not memory
- 115+ indexed tools across 6 backends
- Progressive disclosure:
compass() → describe() → execute() - Tool chains and workflow detection
- Status: ✅ Production • Stack: Python, HNSW, Ollama
Semantic File Search for AI Workstations
- Sub-100ms search across 10k+ file codebases
- Project-aware context boundaries
- Integrates with Tool Compass
- Status: ✅ Production • Stack: Python, HNSW, Ollama
Production-Ready ComfyUI Client
- Headless ComfyUI orchestration
- AI-powered prompt intelligence
- Video generation workflows
- Modern Gradio UI
- Status: ✅ Production • Stack: Python, ComfyUI, Gradio
Headless LLM Fine-Tuning
- Simple Python API with smart defaults
- Multi-run SLAO training
- GGUF export for Ollama/llama.cpp
- Full Windows + WSL2 support
- Status: ✅ Production • Stack: Python, Hugging Face
AI-Powered Code Analysis MCP
- Test generation (unit, integration, e2e)
- Security audits (49 detection patterns)
- Code health scoring across 11 categories
- Multi-language: Python, JS/TS, PowerShell, CSS, HTML
- Status: ✅ Production • Stack: Python, FastMCP
Vector-Embedded Gradio Components
- Semantic codebase navigation
- AI-native UI components
- Context-aware search
- Status: 🚧 Active Development • Stack: Python, Gradio
User Query → Tool Compass (find relevant tools)
→ File Compass (locate implementation)
→ Dev-Brain (analyze code health)
Code Changes → Dev-Brain (security audit + tests)
→ File Compass (find related files)
→ Tool Compass (suggest refactoring tools)
Text Prompt → Comfy-Headless (generate image/video)
→ Tool Compass (find post-processing tools)
→ File Compass (store in asset library)
Training Data → Backpropagate (fine-tune model)
→ Tool Compass (deploy to inference)
→ Dev-Brain (validate outputs)
git clone https://github.com/mcp-tool-shop-org/tool-compass
cd tool-compass
pip install -e .{
"mcpServers": {
"tool-compass": {
"command": "python",
"args": ["-m", "tool_compass"]
}
}
}# Natural language tool discovery
tools = compass("semantic file search")
# Returns: file-compass, integradio, etc.| Principle | Implementation |
|---|---|
| Local-First | All tools run on local infrastructure (RTX 5080 optimized) |
| Semantic Search | HNSW + Ollama embeddings across all layers |
| Production-Ready | Error handling, logging, analytics, monitoring |
| Modular Design | Tools compose via MCP protocol |
| FastMCP Framework | Built on modern Python MCP SDK |
| Tool | Operation | Latency | Scale |
|---|---|---|---|
| Tool Compass | Tool discovery | <20ms | 115+ tools |
| File Compass | File search | <100ms | 10k+ files |
| Comfy-Headless | Image generation | 2-3x speedup | TensorRT |
| Dev-Brain | Code scan | ~50ms/file | 49 patterns |
Core Technologies:
- Model Context Protocol (MCP)
- FastMCP (Python MCP SDK)
- HNSW (hnswlib) for vector search
- Ollama (nomic-embed-text embeddings)
AI Infrastructure:
- ComfyUI (image/video generation)
- TensorRT (inference optimization)
- vLLM (LLM serving)
- Hugging Face (model training)
Frameworks:
- Gradio (web UIs)
- FastAPI (REST APIs)
- SQLite + ChromaDB (persistence)
All repositories include:
- ✅ Comprehensive README with quickstart
- ✅
llms.txtfor AI context (GitMCP compatible) - ✅ Semantic versioning with Git tags
- ✅ GitHub Actions CI/CD
- ✅ Production-grade error handling
- ✅ API/tool reference documentation
For AI Agents:
- Discover and orchestrate 115+ tools by intent
- Navigate codebases semantically, not textually
- Generate test suites and security audits automatically
- Create visual assets on-demand
For Developers:
- Reference architecture for production MCP servers
- Integrate semantic search into applications
- Fine-tune models for specific domains
- Automate code quality workflows
- Claude Desktop - Native MCP client
- VS Code - via MCP extension
- Cursor - AI-native code editor
- Windsurf - Cloud development
- Custom Clients - Any MCP-compatible tool
- MCP Registry submissions for all servers
- Cross-tool workflow automation
- Distributed context mesh (CID-based addressing)
- Enhanced analytics and observability
- Community server templates
We welcome contributions! Each repository has its own contribution guidelines. General principles:
- Syntropy first - Systematic, production-ready solutions
- Comprehensive docs - READMEs, llms.txt, API references
- Error handling - No silent failures
- Tests - Unit, integration, e2e where applicable
- Semantic versioning - Tag releases properly
- Organization: @mcp-tool-shop
- Issues: Use individual repository issue trackers
- Discussions: Coming soon
- Philosophy: Syntropy Manifesto (planned)
Each repository maintains its own license. Check individual repos for details.
Built with 🧠 by developers, for AI agents
Syntropy above all else