I got frustrated with Claude Code memory plugins like claude-mem (buggy, token-heavy, complex setup) and Continuous-Claude (overwrites configs, needs Python deps), so I built a dead-simple alternative: cc-history-search. Zero deps and no setup beyond install.
I've tried it on my existing projects and it works VERY well.
It uses existing chat data directly: no fancy FTS/vector DBs, only simple file searches. Because of that there is no indexing step and no surprise token usage.
Wondering why Anthropic hasn't added it as a first-party feature.