Quick Start
1. Start the system
km start
2. Index your first repository
km index ~/path/to/your/project
Knowledge Master will:
- Parse all code files (Python, TypeScript, Rust, Go, etc.)
- Split them into smart chunks (by function/class boundaries)
- Generate embeddings locally (nomic-embed-text via Ollama)
- Store in the knowledge graph with relationships
- Auto-detect tech stack, services, and conventions
3. Search
km search "how does authentication work"
Results include graph context: which repo, who authored it, related services.
4. Check blast radius
km blast-radius postgres
Shows everything that depends on postgres — repos, services, people.
5. Start the web UI
km serve
Open http://127.0.0.1:9999 for:
- Visual search with results
- Interactive knowledge graph visualization
- File browser for adding new sources
- Source management (add/remove)
6. Connect to AI agents
Knowledge Master works as an MCP server. Your AI agent can search, check blast radius, and enforce conventions directly.
See MCP Integration for setup with Claude, Cursor, or Kiro.