latent-k
Stop wasting tokens - give your AI a map of your code
Start new thread
Jordi Zaragoza

latent-k - Stop wasting tokens - give your AI a map of your code

by•
AI coding assistants waste tokens re-exploring your codebase every session. latent-k fixes this by maintaining a persistent dependency map that's automatically injected as context. - Auto-syncs when files change - ~3k tokens for 80 files (vs ~15k raw) - Tracks imports, exports, and callers - MCP server for on-demand queries - Persistent notes across sessions Works with Claude Code & Gemini CLI. Supports JS/TS, Python, Go, Rust, Java, PHP, Ruby, C#, C/C++. 14-day free trial, then $9/month.

Add a comment

Replies

Best
Jordi Zaragoza
Maker
📌
Hey! I'm Jordi, the maker of latent-k. I built this because I was frustrated with how much context AI coding assistants waste re-exploring code. The idea is simple: give them a map instead of making them wander. Check out the demo video - I ran the same task with vanilla Claude Code vs with latent-k. The funny part: I had time to install latent-k from scratch and STILL finished before vanilla Claude figured out the codebase structure. Happy to answer any questions about how it works under the hood!