About this project
PitchA local-first persistent memory engine for AI coding agents — a single, zero-dependency Go binary that gives your agent a durable knowledge graph of everything it learns across sessions. Instead of re-explaining your conventions, past bugs, and architectural decisions every time, agents save and recall four kinds of memory (session summaries, task patterns, error resolutions, and user rules) through a clean 4-tool MCP bridge or CLI. It ships with a built-in PII-scrubbing pipeline that strips secrets and credentials before anything touches disk, SQLite FTS5 full-text search with BM25 + Jaccard near-duplicate detection so the same lesson is never stored twice, and a privacy-by-default design: everything stays on your machine, loopback-bound, with no external APIs and no CGO. The two-tier context bundle surfaces the right memories at the right depth — always-on rules and summaries up front, browsable patterns and fixes behind them — so agents start every task already knowing your codebase. Fast, auditable, and frictionless (auto-spawning server, no config required), droids-mem turns stateless agents into ones that actually remember.








Discussion
0 commentsSign in to join the discussion.
No takes yet. Be the first to weigh in.