Claude Code not indexing properly?

[eja-git]

Hey J, hopefully this is the right forum for this. I've been trying to get Claude Code setup to use jcodemunch-mcp properly and it's almost there, but it doesn't seem to be reading from ~/.code-index at the start of a new session? Is that something I need to force? I also have gemini and codex clis installed, they must be loading every repo in ~/.code-index every session, because I can list_repos and I see all the indexed repos.

I asked claude.ai to help me work through the problem, here's what it said:

"Better fix: This seems like something worth raising as an issue on the jCodeMunch repo. The expected behavior should be that list_repos scans ~/.code-index/ and returns everything cached on disk, not just what was indexed during the current process lifetime. The other CLIs might just be longer-lived processes, or they might be calling list_repos in a way that triggers a cache scan.
You could also try asking Claude Code to run invalidate_cache and then list_repos again — sometimes that forces a rescan. But I'd bet the real answer is filing an issue with the jCodeMunch author about cross-process index discovery."

Anyway, any thoughts appreciated, thanks!

[DrHayt]

This is almost certainly that claude has the mcp server available but hasn't been instructed to use it.

uvx jcodemunch-mcp config --check

Will show you if it is configured properly, and if it isn't, you can always:

uvx jcodemunch-mcp init

Of course this assumes you run your stuff under uvx, you dont have to, but I do :)