Excessive memory usage

[wuxin-gh]

I have to raise this question to you because my claude code CLI started multiple times, and my system memory exploded. You can see this is the usage of my claude code, almost every process takes up more than 200 MB. I started it on the terminal, installed it on npm, and the system is win. The machine has 14 cores and 32GB. However, since using yours, my computer's fan has been running non-stop, and every time the system memory is occupied, it takes up about 80%. My claude code is the latest version updated yesterday, and the version of codegraph is 1.0.1

[colbymchenry]

Thanks for the detailed report and the screenshot — that helps a lot.

The key thing: you're on CodeGraph 1.0.1, and the current release is 1.1.6 (six releases later). Almost everything you're seeing was fixed in that window:

• Processes piling up — 1.0.1 had a bug where a stale lock could let duplicate background servers accumulate for the same project instead of sharing one. Fixed in 1.1.1 (Codex MCP tool call fails with Transport closed on v1.1.0 while CLI status/sync are healthy #974). Since then it's one shared server per project across all your sessions, and it shuts itself down after 5 minutes idle.
• Fan / constant CPU — on larger repos 1.0.1 could pin a CPU core for many minutes in reference resolution. Fixed in 1.1.0 (Under large projects, using codegraph index or codegraph init -t results in extremely slow retrieval speeds #915) and 1.1.2 (codegraph index wedges indefinitely at "Resolving refs" 94% on a repo with a large *tracked* JS/TS asset dir and there's no way to exclude it #999).
• Memory / index bloat — 1.0.1 could index into gitignored folders of nested repos and build a huge graph. Fixed in 1.1.1 ([BUG] Currently, it is impossible to turn off the indexing of nested repositories #976) and 1.1.4 (v1.1.3 still indexes gitignored embedded repos; ignored benchmark corpus triggers resolving-refs watchdog #1065).

Please upgrade to 1.1.6 (npm i -g @colbymchenry/codegraph@latest, or re-run the installer) and let me know if it persists.

Two honest notes from your screenshot, so you know where the memory is actually going: several of those ~200 MB "Claude Code" processes are Claude Code's own CLI — one per session you launch, independent of CodeGraph — and the single largest item, "Orca" at ~1.4 GB, isn't CodeGraph at all. CodeGraph's own processes there are the smaller 40–56 MB ones.

On a 14-core machine, if you'd like an even lighter footprint after upgrading, you can set CODEGRAPH_QUERY_POOL_SIZE=2 and CODEGRAPH_PARSE_WORKERS=4 to cap how many helper threads it uses. Happy to help tune further if needed.