# Does Claude Code need some kind of persistent map of a codebase? — retained record

> Retained reddit record from reddit:ClaudeCode.

Canonical: https://highsignal.app/data/records/067cffd525c463cc

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeCode)  published 2026-09-14 16:19 ingested 2026-09-15 02:40

[original source ↗](https://www.reddit.com/r/ClaudeCode/comments/1wg80fa/does_claude_code_need_some_kind_of_persistent_map/)

## retained text



```
I’ve been using Claude Code a lot on bigger repos and one thing I’ve noticed is how much time it spends figuring out stuff it has basically already figured out before.  
Like where things live, what calls what, which files are related, how a system is structured, etc. Then you start a new session or compact and it has to work a lot of that out again.  
CLAUDE.md definitely helps but I don’t think it completely solves it, especially once a repo gets pretty large.  
I work on an open source project called Ix that builds a persistent graph of a codebase and lets Claude query it, so I’m obviously a little biased here. But working on it has made me curious whether this is actually something other people run into or if we’re overengineering the problem.  
For people using Claude Code on bigger codebases, have you noticed much of a difference using things like repo maps, indexing, RAG, knowledge graphs, etc. compared to just letting Claude search the repo normally?  
I’m starting to think the useful part isn’t really “memory”, it’s just giving Claude a cheap way to answer stuff like “what depends on this?” or “where does this system start?” without reading a bunch of files first.  
Curious what other people have found.
```



document fetched 2026-09-15 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1wg80fa/does_claude_code_need_some_kind_of_persistent_map
