Compatibility

Works with the agent you already use

deck works with any coding agent that can run a shell command, including Claude Code, Codex, Cursor and Amp. No new model, no API key, no account.

deck does not run a model, and it does not replace your agent. The agent you already use writes the deck, in the same session, with everything it just read.

Any agent with a shell

A deck is written with four shell commands: deck new, deck open, deck group and deck seal. Any agent that can run a command can use it. That includes:

• Claude Code
• Codex
• Cursor
• Amp

deck setup finds the agents on your machine and installs a short skill into each, so they know when a deck is a better answer than a paragraph.

The catch, for agents with reply hooks

Some agents can run a hook when they finish a reply. For those, setup offers the catch: when a reply names two or more file:line locations, it is sent back to the agent to become a deck instead. Claude Code supports this today. Agents without hooks still use deck whenever their skill suggests it.

What you need

• macOS, Linux or Windows
• one of the agents above, or any agent with a shell
• nothing else: no account, no API key, no paid tier

Install it from the home page, or read the docs.