← ClaudeAtlas

rememlisted

Use when the user asks Codex to recall prior project context, save durable decisions or bug fixes, inspect remem memory health, or activate remem automatic memory hooks from the Codex plugin.
majiayu000/remem · ★ 30 · AI & Automation · score 79
Install: claude install-skill majiayu000/remem
# Remem Remem gives Codex durable project memory through the `remem` MCP server. Use it when the answer depends on prior sessions, project preferences, saved decisions, workstreams, or memory health. ## Retrieval - Search remem before answering questions about prior project context, remembered preferences, older bug fixes, workstreams, or "what did we decide before". - Use compact `search` results first, then call `get_observations` for selected IDs when exact content is needed. - Use `search_raw` only when curated search misses an exact phrase or chat transcript detail. - Use experimental `context_bundle` with `schema_version: 1` when a caller needs one policy-bounded, source-attributed SessionStart bundle plus its complete selection/drop audit. Its wire shape is versioned but not yet stable, and the poisoning safety check may quarantine unsafe persisted rows while retaining a redacted audit identity. Poisoned session/workstream text is removed before implicit retrieval query construction. Canonical memory, session, and preference preselection drops retain their stable identity and exact audit reason. Use `risk: high` only when trusted user-authored memory is sufficient; the compiler abstains if no trusted item survives. The foreground compiler never calls a configured remote embedding provider and does not apply ambient reranker configuration outside the versioned plan. - Treat memory as evidence, not current truth. Verify live repo, GitHub, filesys