claude-mem vs operator-memory
Cross-harness session memory: hooks capture what the agent does, an LLM compresses it into observations, and the next session gets the relevant ones back via progressive-disclosure MCP tools. — versus — Context engine for coding agents: each session consults, then updates, a Markdown brain of specs, decisions, standards and lessons — private, team-shared via Git, and per-user. No embeddings.
Both give coding agents memory across sessions. claude-mem captures and compresses what the agent did into retrievable observations; Operator rejects snippet capture and has the agent edit canonical spec documents instead.
| claude-mem | operator-memory | |
|---|---|---|
| Stars | 113k | 500 |
| Forks | 9.8k | 19 |
| Language | JavaScript | TypeScript |
| License | Apache-2.0 | BSD-3-Clause |
| Last activity | yesterday | yesterday |
| Topics | memory, coding | memory, coding |
| Curated connections | 12 | 5 |
claude-mem — the curator's take
The most widely installed of the capture-and-reinject memory plugins, and the one with the most product around it: 5 lifecycle hooks, a Bun worker with an HTTP API and live web viewer, SQLite plus FTS5 plus Chroma for hybrid search, and a 3-layer MCP search pattern (index → timeline → full detail) that keeps recall at ~50-100 tokens per hit until you actually want the body. Installs into Claude Code, OpenCode, Antigravity, OpenClaw and more, with `<private>` tags to keep things out of the store and optional cloud sync. The tradeoffs: it spends model tokens summarizing every session, it drags in Bun and uv, and what you get back is a generated observation rather than the raw transcript — if you want auditable, greppable memory you can edit by hand, look at the Markdown-first options instead.
operator-memory — the curator's take
Use it when you want agent memory that looks like project documentation: specs and decisions the agent keeps current, readable in a diff and shareable with your team by committing .operator-shared/. Works across Claude Code, Codex, OpenCode, Pi and DeepSeek's harness with no background service or vector DB. NOT a recall store for chat transcripts or fuzzy retrieval — a cold project starts with a sparse brain until you ask for the first specs, and it relies on the agent choosing to update docs, so long sessions can drift.