Every prompt saved. Zero effort.

Use Claude Code, Codex, or Gemini like you always do. Prompt Cellar captures every prompt in the background—encrypted, timestamped, and ready to find later.

How it works

One install command. Then forget about it.

1. Install the CLI hook

Run pc setup once. It installs a lightweight hook for your AI tools. Takes about 30 seconds.

2. Use your tools normally

Open Claude Code, Codex, or Gemini. Send prompts. Work like you always do. The hook runs silently in the background.

3. Find any prompt later

Open the Prompt Cellar dashboard. Every prompt is there, grouped by session and project. Search by keyword, filter by tool, or browse your timeline.

What gets captured

Not just the prompt text. The full context around it.

Prompt text

The full text of every prompt you send, encrypted before it leaves your machine.

Project context

Repository name, git branch, and working directory. Know exactly where each prompt was used.

Session grouping

Prompts from the same CLI session are grouped together, so you can replay an entire coding session.

Tool & model

Which tool you used (Claude Code, Codex, Gemini) and which model was active.

Timestamp

Exact time each prompt was sent. See when you worked on what.

Environment

MCP servers, hooks, skills, and config files active during the session.

Frequently asked questions

How does auto-capture work?

Prompt Cellar installs a lightweight CLI hook that runs in the background. When you send a prompt to Claude Code, Codex, or Gemini, the hook captures the text and metadata before it leaves your machine. You don't change anything about how you work.

Does auto-capture slow down my CLI tools?

No. The capture hook adds negligible overhead — a few milliseconds at most. Your prompt is sent to the AI tool at the same time it's captured. There's no waiting.

What metadata is captured alongside the prompt?

Each prompt is stored with its timestamp, the tool you used (Claude Code, Codex, Gemini), your working directory, git branch, session ID, and repository name. This context makes prompts searchable and browsable later.

Can I turn off auto-capture temporarily?

Yes. You can disable the capture hook at any time with a single command. When you re-enable it, capture resumes from where you left off. No prompts are lost during the gap — they simply aren't captured.

Is my prompt data encrypted?

Yes. Every captured prompt is encrypted client-side with AES-256-GCM before it's sent to the server. The server stores encrypted blobs and cannot read your prompt content.

Which AI coding tools are supported?

Prompt Cellar supports Claude Code (via native hook), Codex CLI (via shell wrapper), and Gemini CLI (via shell wrapper). All three tools' prompts appear in a single unified timeline.

Start capturing every prompt

One command to set up. First 100 prompts free.

Get Started Free