Hey There! I built a companion package for agent-shell called agent-recall and wanted to share it.
What it does
Currently, there's no main repository for ALL agent-shell chats across projects, which makes it painful and tedious to hunt down specific past conversations.
Agent-Recall indexes all of your transcripts, provides full-text search, and lets you resume past conversations directly from any transcript. Session resume currently supports Claude CLI only, but search and browse work with any agent-shell transcript. Open to PRs for other agents!
- Full-text search across all transcripts (powered by ripgrep)
- Live search with real-time filtering via counsel-rg or consult-ripgrep
- Browse transcripts organized by project and date (can set different ordering)
- Resume past agent-shell conversations from any transcript (Currently only Claude CLI)
- Persistent index that grows automatically via a mode hook
- Stats overview of your transcript collection
Quick demo
;; Point it at your project roots
(setq agent-recall-search-paths '("~/projects" "~/work"))
;; Build the index once
M-x agent-recall-reindex
;; Then search, browse, or resume
M-x agent-recall-search
M-x agent-recall-browse
M-x agent-recall-resume
How it relates to agent-shell
It reads the .agent-shell/transcripts/ Markdown files that agent-shell produces — no modifications to agent-shell needed. It also hooks into agent-shell-mode-hook to track session IDs for instant resume.
Repo: https://github.com/Marx-A00/agent-recall
Thought this might be useful to other agent-shell users. Happy to hear any feedback!
Hey There! I built a companion package for agent-shell called agent-recall and wanted to share it.
What it does
Currently, there's no main repository for ALL agent-shell chats across projects, which makes it painful and tedious to hunt down specific past conversations.
Agent-Recall indexes all of your transcripts, provides full-text search, and lets you resume past conversations directly from any transcript. Session resume currently supports Claude CLI only, but search and browse work with any agent-shell transcript. Open to PRs for other agents!
Quick demo
How it relates to agent-shell
It reads the
.agent-shell/transcripts/Markdown files that agent-shell produces — no modifications to agent-shell needed. It also hooks intoagent-shell-mode-hookto track session IDs for instant resume.Repo: https://github.com/Marx-A00/agent-recall
Thought this might be useful to other agent-shell users. Happy to hear any feedback!