recall · arjunkmrm — Claude Code skill: search past agent sessions
Full-text search across past Claude Code, Codex, pi and Grok sessions with BM25 ranking. Useful when you need to remember what you already discussed with an agent.
npx skills add arjunkmrm/recall --skill recall -a claude-code
npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a claude-codeWhat recall does
The skill indexes four sources: ~/.claude/projects/ for Claude Code, ~/.codex/sessions/ for Codex, ~/.pi/agent/sessions/ for pi, and ~/.grok/sessions/**/chat_history.jsonl for Grok. Only user and assistant messages make it into the index — tool calls, thinking blocks and state snapshots are skipped. The database lives at ~/.recall.db on SQLite FTS5.
Queries follow FTS5 syntax: exact phrases in quotes, AND, OR, NOT, and prefix matching with an asterisk. Japanese, Chinese and Korean text falls back to trigram matching, English uses Porter stemming. You can also skip the query entirely and get a plain list of sessions in the window, sorted by recency.
The first run indexes everything in a few seconds; later runs are incremental. Upgrading from 0.3.x, 0.4.x or 0.2.x means running --reindex once, otherwise the newer sources and the CJK index stay missing. Python 3 is required — the script runs as python3 ~/.claude/skills/recall/scripts/recall.py.
What it helps with
- recall which approach you settled on last session
- list sessions for one project within a date window
- narrow history down to Codex or pi only
- read the raw transcript of a matched session
- resume a session with
--resumeor--session
Example prompts for the agent
- search my past sessions for what we discussed about bufferStore
- list every pi session from the last week
- find the exact phrase "ACP protocol" in my Claude Code history
How to install recall
1Skills CLI (Vercel)
npx skills add arjunkmrm/recall --skill recall -a claude-codenpx skills add arjunkmrm/recall --skill arjunkmrm-recall -a claude-code2Manually: put the files in place
~/.claude/skills/arjunkmrm-recall/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/arjunkmrm-recall/
1Skills CLI (Vercel)
npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a codex2Manually: put the files in place
.agents/skills/arjunkmrm-recall/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a cursor2Manually: put the files in place
.cursor/skills/arjunkmrm-recall/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install arjunkmrm/recall arjunkmrm-recall2Manually: put the files in place
.github/skills/arjunkmrm-recall/SKILL.md1Skills CLI (Vercel)
npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a gemini-cli2Platform CLI
gemini skills install https://github.com/arjunkmrm/recallManually: put the files in place
.devin/skills/arjunkmrm-recall/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a opencodeClaude.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
How do I install recall in Claude Code?
Run npx skills add arjunkmrm/recall --skill recall -a claude-code, then npx skills add arjunkmrm/recall --skill arjunkmrm-recall -a claude-code. The skill file lands at ~/.claude/skills/arjunkmrm-recall/SKILL.md.
What does recall do?
It searches past Claude Code, Codex, pi and Grok sessions using full-text search with BM25 ranking. Results carry [claude], [codex], [pi] or [grok] tags.
How does English search differ from CJK?
The index has two tables: English queries go through Porter stemming, while Japanese, Chinese and Korean text uses trigram matching.
Source text
SKILL.md · MIT
--- name: recall description: > Search past Claude Code, Codex, pi and Grok sessions. Triggers: /recall, "search old conversations", "find a past session", "recall a previous conversation", "search session history", "what did we discuss", "remember when we" metadata: author: arjunkmrm version: "0.4.1" license: MIT --- # /recall — Search Past Claude, Codex, pi & Grok Sessions Search all past Claude Code, Codex, pi and Grok sessions using full-text search with BM25 ranking. ## Usage ```bash python3 ~/.claude/skills/recall/scripts/recall.py [QUERY] [--project PATH] [--days N] [--source claude|codex|pi|grok] [--limit N] [--reindex] ``` ## Examples ```bash # List every session in the last day (no text search) python3 ~/.claude/skills/recall/scripts/recall.py --days 1 # List every pi session in the last week python3 ~/.claude/skills/recall/scripts/recall.py --days 7 --source pi # Simple keyword search python3 ~/.claude/skills/recall/scripts/recall.py "bufferStore" # Phrase search (exact match) python3 ~/.claude/skills/recall/scripts/recall.py '"ACP protocol"' # Boolean query python3 ~/.claude/skills/recall/scripts/recall.py "rust AND async" # Prefix search python3 ~/.claude/skills/recall/scripts/recall.py "buffer*" # Filter by project and recency python3 ~/.claude/skills/recall/scripts/recall.py "state machine" --project ~/my-project --days 7 # Search only Claude Code sessions python3 ~/.claude/skills/recall/scripts/recall.py "buffer" --source claude # Search only Codex sessions python3 ~/.claude/skills/recall/scripts/recall.py "buffer" --source codex # Search only pi sessions python3 ~/.claude/skills/recall/scripts/recall.py "buffer" --source pi # Search only Grok sessions python3 ~/.claude/skills/recall/scripts/recall.py "buffer" --source grok # Force reindex python3 ~/.claude/skills/recall/scripts/recall.py --reindex "test" ``` ## Query Syntax (FTS5) - **Words**: `bufferStore` — matches stemmed variants (e.g., "discussing" matches "discuss") - **Phrases**: `"ACP protocol"` — exact phrase match - **Boolean**: `rust AND async`, `tauri OR electron`, `NOT deprecated` - **Prefix**: `buffer*` — matches bufferStore, bufferMap, etc. - **Combined**: `"state machine" AND test` - **CJK**: `issue化` — automatically uses trigram matching for Japanese/Chinese/Korean text ## After Finding a Match To resume a session, `cd` into the project directory and use the appropriate command: ```bash # Claude Code sessions [claude] cd /path/to/project claude --resume SESSION_ID # Codex sessions [codex] cd /path/to/project codex resume SESSION_ID # Grok sessions [grok] ```bash grok --resume SESSION_ID ``` # Pi sessions [pi] cd /path/to/project pi --session SESSION_ID # full or partial id; pi resolves prefix matches ``` Each result includes a `File:` path. Use it to read the raw transcript (auto-detects format): ```bash python3 ~/.claude/skills/recall/scripts/read_session.py <File-path-from-result> ``` If results are missing `File:` paths, run `--reindex` to backfill. ## Notes - Index is stored at `~/.recall.db` (SQLite FTS5, auto-migrated from `~/.claude/recall.db`) - Indexes four sources: `~/.claude/projects/` (Claude Code), `~/.codex/sessions/` (Codex), `~/.pi/agent/sessions/` (pi), and `~/.grok/sessions/**/chat_history.jsonl` (Grok) - First run indexes all sessions (a few seconds); subsequent runs are incremental - Only user and assistant messages are indexed (tool calls, thinking blocks, state snapshots skipped) - Results show `[claude]`, `[codex]`, `[pi]`, or `[grok]` tags to indicate the source - Dual-table FTS: English queries use Porter stemming, CJK queries use trigram matching - Omit the query argument for **list mode** — every session in the window, sorted by recency, no FTS - Provide a query for full-text search; both modes accept `--project`, `--days`, `--source`, `--limit` - **Upgrading from 0.3.x**: run `--reindex` once to pull in pi sessions - **Upgrading from 0.4.x**: run `--reindex` once to pull in Grok sessions - **Upgrading from 0.2.x**: run `--reindex` once to build the CJK index
© arjunkmrm · MIT · arjunkmrm/recall
Similar skills
All skills for Agent setup: memory, context, orchestration →recall by other authors comes first.
