MCP Knowledge Service

MCP-based knowledge and rules suite for Tailscale networks with semantic search via libSQL vectors.

Features

• MCP Tools: rules.search, project.search, refs.list, and more

• Semantic Search: Vector-based search using libSQL with OpenAI embeddings

• Multi-Channel: Support for multiple MCP channels (rules, projects, refs)

• Tailscale Security: Identity-based access control via Tailscale Serve

• Vector Database: libSQL with native vector support for ANN search

Quick Start

1. Setup Environment:

   ./scripts/setup.sh

2. Configure Environment: Edit .env with your configuration:

   LIBSQL_URL=file:./data/knowledge.db
   OPENAI_API_KEY=your-openai-api-key

3. Development:

   npm run dev        # Start development server
   npm run build      # Build for production
   npm test          # Run tests
   npm run lint      # Lint code

Architecture

• src/mcp/ - MCP server and tool implementations

• src/db/ - Database schema, connections, and migrations

• src/http/ - REST API endpoints for ingestion

• src/auth/ - Tailscale identity and access control

• src/utils/ - Shared utilities and helpers

MCP Tools

Rules Service

• rules.search(q, k?, tags?) - Search global development rules

• rules.get(id) - Get specific rule by ID

• rules.tags() - List all available rule tags

Project Service

• project.search(project, q, k?) - Search within project docs

• project.browse(project, path?) - Browse project structure

• project.contextPack(project, facets?) - Get curated context bundle

References Service

• refs.list(tags?, limit?) - List references with optional tag filter

• refs.add(title, url, note?, tags_csv?) - Add new reference

• refs.findByTag(tag) - Find references by specific tag

Database Schema

The service uses libSQL with vector support:

• rules_global - Global AI development rules with embeddings

• project_docs - Project-specific documentation with embeddings

• refs - Quick reference links and documentation

• access_tiers - User access control and permissions

• audit_log - Query audit trail and metrics

Development Status

See TODO.md for current development phases and tasks.

Related

• Memory subsystem development: /home/ubuntu/mem

• Design documentation: docs/memory-design.md