recall-conversations — Claude Code skill: search past sessions
recall-conversations reads back your earlier Claude sessions: finds a decision, restores context, or builds a retro from the local conversation database.
no license ★ 276 updated 2026-10-29
/plugin marketplace add gupsammy/Claudest
/plugin install claude-memory@ClaudestWhat recall-conversations does
The skill gives Claude a way back into earlier sessions. Three Python scripts sit on top of the local conversation store. recent_chats.py lists sessions by recency, with --timeline for a one-line index and --session for a named session by uuid or an 8+ character prefix. search_conversations.py does keyword search, ranked by BM25 when FTS5 is present. list_projects.py maps project names to paths and session counts.
Each request goes to a lens — restore context, look up one decision, run a retrospective, collect rules you stated but never saved to CLAUDE.md, spot repeated mistakes. Every lens ships a recipe with its own flags, and the questions to ask of the retrieved sessions live in references/lenses.md. Scope defaults to the current project via CWD; --project and --all-projects widen it.
Matching is lexical. A prompt with no concrete terms returns noise, so recency listing beats search there. Broad pulls print whole transcripts and fill the context window; the wide recipes pass --summary for that reason, and when even that is too much you shard by project and hand each shard to a subagent. Scripts run under python3, with paths built from ${CLAUDE_PLUGIN_ROOT}.
What it helps with
- pick up a project where you left it
- find the session where a bug got fixed
- write a retro covering several weeks of work
- turn rules said out loud in chat into
CLAUDE.mdentries - spot mistakes you keep repeating
Example prompts for the agent
- Where did we leave off? Restore context for this project.
- Find the session where we dealt with the login bug and show the matching passages.
- Run a retrospective over the last sessions: what went well, what hurt.
How to install recall-conversations
1Plugin marketplace
/plugin marketplace add gupsammy/Claudest/plugin install claude-memory@ClaudestThis installs the whole claude-memory plugin with all its skills
2Skills CLI (Vercel)
npx skills add gupsammy/Claudest --skill recall-conversations -a claude-code3Manually: put the files in place
~/.claude/skills/recall-conversations/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/recall-conversations/
1Skills CLI (Vercel)
npx skills add gupsammy/Claudest --skill recall-conversations -a codex2Manually: put the files in place
.agents/skills/recall-conversations/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add gupsammy/Claudest --skill recall-conversations -a cursor2Manually: put the files in place
.cursor/skills/recall-conversations/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install gupsammy/Claudest plugins/claude-memory/skills/recall-conversations2Manually: put the files in place
.github/skills/recall-conversations/SKILL.md1Skills CLI (Vercel)
npx skills add gupsammy/Claudest --skill recall-conversations -a gemini-cli2Platform CLI
gemini skills install https://github.com/gupsammy/Claudest --path plugins/claude-memory/skills/recall-conversationsManually: put the files in place
.devin/skills/recall-conversations/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about recall-conversations
How do I install recall-conversations in Claude Code?
Add the marketplace first: /plugin marketplace add gupsammy/Claudest. Then run /plugin install claude-memory@Claudest. The skill arrives as part of that plugin.
What does recall-conversations do?
It searches and rereads your earlier Claude sessions, then restores context or analyses them: retros, decisions, recurring mistakes. Three Python scripts read the local database.
When does search catch paraphrases?
When the search metadata reports retrieval: hybrid, meaning-based matches are fused with keyword hits. Passing --lexical keeps keywords only.