Launched this week
Nomik
Your AI agent's memory. A knowledge graph of your codebase
3 followers
Your AI agent's memory. A knowledge graph of your codebase
3 followers
Stop feeding raw files to your AI agent. Nomik parses your entire codebase into a structured knowledge graph: functions, call chains, routes, DB tables, APIs, events, all connected with real relationships. It plugs into Cursor, Windsurf, and Claude via MCP so your agent queries precise context instead of guessing from file dumps. Ask "what breaks if I change this?" and get a real answer. Supports TypeScript, Python, Rust, SQL, and config files. Free, open-source, 100% local. No cloud, no lock-in
Hey Product Hunt! I'm Hamza, the maker behind Nomik.
I built this because I was frustrated with how AI coding agents handle large codebases. They dump raw files into context, hallucinate architecture, and miss cross-file relationships. I wanted my agent to actually understand the code and not just read it.
Nomik solves this by building a knowledge graph (Neo4j) from your source code using deep static analysis (Tree-sitter). Every function, class, route, DB table, API call, and event gets extracted as a node with typed edges (CALLS, DEPENDS_ON, WRITES_TO, HANDLES, etc.). Then it connects to your AI editor via MCP. So instead of 50 files of noise, your agent gets 5 precise graph nodes.
What you can do today:
Ask your AI "what breaks if I change X?" → real blast radius from the graph
"What calls this function?" → exact callers, not grep guesses
PR impact analysis, dead code detection, architecture drift between commits
Full codebase onboarding in seconds
Works with Cursor, Windsurf, Claude Desktop
many more stuff.
It's 100% local (Neo4j on your machine), free, and open-source under FSL-1.1.
I'd love feedback, especially on what node/edge types or languages you'd want supported next.
Thanks for checking it out!