English | Chinese

Recall

CodSpeed

Local-first search across every AI coding session on your machine.

Recall

Jump between Claude Code, Codex, and whatever comes next; Recall pulls those scattered local sessions into one searchable index, tracks usage when token metadata is available, and drops you back into the original CLI.

Architecture

Recall architecture

Install

brew install samzong/tap/recall

Usage

recall sync          # incremental sync of the current project (safe to run anytime)
recall sync --project all  # sync every project and run index-wide maintenance
recall               # launch TUI
recall usage         # usage dashboard
recall export --project all > recall-export.jsonl # export all sessions
recall import recall-export.jsonl --dry-run  # preview an import
recall session list  # list sessions for agents/scripts
recall session share --id <session-id> --format json  # publish one selected session
recall share list --format json  # list published share URLs
recall share unpublish <share-id> --yes  # take down one public page
recall info  # index stats and worker status

With Skill use Recall is the best way.

recall skill install # auto detect agents and install skills

Support

One index across every AI coding CLI. Sync once, search everywhere, resume right where you left off.

Adapter Discovery Full-index Incremental-sync Semantic-search Export Resume Usage
Claude Code ✅ ✅ ✅ ✅ ✅ ✅ ✅
OpenCode ✅ ✅ ✅ ✅ ✅ ✅ ✅
Codex ✅ ✅ ✅ ✅ ✅ ✅ ✅
Pi ✅ ✅ ✅ ✅ ✅ ✅ ✅
OMP ✅ ✅ ✅ ✅ ✅ ✅ ✅
Antigravity ✅ ✅ ✅ ✅ ✅ ✅
Gemini ✅ ✅ ✅ ✅ ✅ ✅ ✅
Kiro ✅ ✅ ✅ ✅ ✅ —
Copilot ✅ ✅ ✅ ✅ ✅ ✅ ✅
Copilot Chat ✅ ✅ ✅ ✅ ✅ —
Cursor ✅ ✅ ✅ ✅ ✅ — ✅
Cline ✅ ✅ ✅ ✅ ✅ —
Roo ✅ ✅ ✅ ✅ ✅ —
DeepSeek Harness ✅ ✅ ✅ ✅ ✅ — ✅
Grok ✅ ✅ ✅ ✅ ✅ ✅ ✅
Kimi Code ✅ ✅ ✅ ✅ ✅ ✅ ✅
Qwen Code ✅ ✅ ✅ ✅ ✅ ✅ ✅
Kilo Code ✅ ✅ ✅ ✅ ✅ ✅ ✅
Crush ✅ ✅ ✅ ✅ ✅ ✅ ✅
MiMo Code ✅ ✅ ✅ ✅ ✅ ✅ ✅
ZCode ✅ ✅ ✅ ✅ ✅ ✅ ✅
Goose ✅ ✅ ✅ ✅ ✅ ✅ ✅

Acknowledgements

  • Thanks to tokscale for the usage dashboard reference and token accounting behavior.
  • Thanks to Ratatui and Crossterm for the terminal UI foundation.
  • Thanks to sqlite-vec and SQLite FTS5 for keeping local text and vector search embedded.
  • Thanks to Candle, Hugging Face, and intfloat/multilingual-e5-small for local semantic embeddings.
  • Thanks to kitup for the bundled agent skill installer.
  • Thanks to LINUX DO for the open-source sharing community.

License

This project is licensed under the MIT License.