Sense a local code intelligence for AI coding tools (open source, looking for testers)
Hey makers! I've been building with Claude Code for months and one thing keeps bugging me. Every time it needs to understand your codebase, it greps around, reads files one by one, and pieces things together blind. It works, but it's slow and sometimes misses the bigger picture.
I know the space is crowded. There are plenty of tools tackling this now and I've tested most of them. Graphify, Gortex, code-review-graph, tokensave, Repomix, Drift, codebase-memory-mcp... My daily driver before this was GrepAI. They all do useful things.
But I kept wanting one tool that covered the full picture without needing API keys, a Python runtime, or a cloud service. So I built Sense.
Four capabilities in one binary:
Graph. Who calls this function? What does it call? Where are its tests? Similar to codebase-memory-mcp or Gortex, but with embedded semantic search.
Search. Find code by meaning, not strings. "Payment error handling" finds the right files even if those words don't appear in the source. Think what Graphify or Cursor do, but local and bundled. No API key, no Ollama.
Blast. What breaks if I change this? Traces callers, flags affected tests, gives a risk score. Like code-review-graph or Engram, but structural rather than statistical.
Conventions. What patterns does this project follow? Only Drift really does this. It's the least common capability in the space and honestly the one I find most useful day to day.
Single Go binary. No API keys, no cloud, no Python/Node runtime. SQLite index stays on your machine. Works with anything that speaks MCP.
Why I'm posting here. I've been running it on my own projects and it works well for me, but I need it tested on codebases that aren't mine. Looking for a handful of early testers willing to point it at a real project and tell me what's broken, confusing, or useless. It's pre-alpha, rough edges expected.
-> GitHub: https://luuuc.github.io/sense/
Happy to answer any questions.
Replies