@chrisdburr/tea-techniques-mcp

MCP server for the TEA Techniques knowledge graph — discover AI assurance techniques through natural language.

Quick Start

No installation needed — run directly with npx:

npx @chrisdburr/tea-techniques-mcp

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "tea-techniques": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@chrisdburr/tea-techniques-mcp"]
    }
  }
}

Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "tea-techniques": {
      "command": "npx",
      "args": ["-y", "@chrisdburr/tea-techniques-mcp"]
    }
  }
}

After configuring, restart Claude. You should see tools prefixed with mcp__tea-techniques__.

Features

• 10 tools for searching, filtering, comparing, and exploring AI assurance techniques
• Semantic search — embedding-based claim matching with hybrid RRF ranking
• Knowledge graph with 92 techniques, 7 assurance goals, and 450+ academic resources
• Zero configuration — fetches data remotely from GitHub Pages with 24h caching

How It Works

The server fetches the TEA Techniques knowledge graph from GitHub Pages on first run and caches it locally for 24 hours (~/.cache/tea-techniques-mcp/). Semantic search uses a lightweight ONNX model (~30MB, downloaded on first use) to match natural-language claims against technique embeddings.

Tool Reference

| Tool | Description | |------|-------------| | find_techniques | Search and filter techniques by query, goals, tags, complexity | | get_technique | Get full details of a specific technique by slug | | compare_techniques | Side-by-side comparison of 2-5 techniques | | find_related | Find related techniques via links, shared goals, or tags | | suggest_techniques_for_claim | Match assurance claims to relevant techniques using semantic search | | find_evidence_types | Explore what evidence types techniques can produce | | explore_taxonomy | Navigate the hierarchical tag taxonomy | | coverage_statistics | Analyse dataset coverage across dimensions | | search_resources | Search academic papers, software, and documentation | | get_knowledge_graph_summary | High-level statistics about the knowledge graph |

Development

This package is part of the TEA Techniques monorepo. For local development:

git clone https://github.com/alan-turing-institute/tea-techniques.git
cd tea-techniques/mcp-server
pnpm install
pnpm dev          # Run from source with tsx
pnpm dev --local  # Load data from local project files

License

MIT