deck
Stop approving code and decisions you don't understand
curl -fsSL https://raw.githubusercontent.com/henit-chobisa/deck/main/install | shDetails
- Follow on
- Categories
- Use Cases
- Target Audience
- Pricing
- Free
- Featured in
Discovery signals
How AI and people discover deck on PeerPush
About deck
AI agents write code faster than anyone can read it, and we merge what we don't understand. deck is a desktop window your coding agent opens to show you its work: each decision as a plain claim, with the exact lines it rests on lit beside it, one group at a time. You read it, argue with it, and comment on the line itself. The agent answers right there in the deck. The questions under each group are the ones you were about to ask, and they change as you ask them. Nothing is decided for you: you decide, then submit your review back to the agent. Works with Claude Code, Codex, Cursor and Amp, through a skill deck installs and keeps up to date with every upgrade. macOS, Linux and Windows. Free and open source (Apache-2.0). Usage counts are shared only if you say yes.
Product Video
Watch a video demo of deck.
Screenshots
Reviews (4)
Average 4.8 out of 5
Based on 4 reviews
A clear approach to reviewing AI-written code: each agent decision is shown as a plain claim beside the exact lines it rests on, with inline discussion. Free and open source.
Comments (5)
Showing each agent decision as a plain claim next to the exact lines it depends on is a smart way to make AI code review readable. Nice that it works with Claude Code, Codex, Cursor and Amp.
Linking each claim to exact lines makes the review easier to challenge. If the agent edits those lines after I start reviewing, are the affected claims marked stale before I submit?
@GeneralManager_Getatext if the claims are affected by your change, those will be marked stale.
This one hits close. I build everything with Claude Code and I'm not a coder, so I approve a lot I can't really read lol. Plain claims next to the exact lines is exactly what I'd want. Works on Windows too, nice.
@fermentory ✌🏼
Commenting on the exact line and getting the agent reply inside the deck beats scrolling a long diff. When I push back on a claim and the agent changes the code, does the deck regroup the decisions or append a new group?
@stackshortlist It doesn’t need to create a new group for you. You’re asking questions based on the current group. Deck allows the agent to open panes, on the fly too. So in response to your comments, it can also show you the exact thing.
Showing each decision as a plain claim with the exact lines it rests on is a smart answer to merging code nobody fully read. Arguing on the line and getting the agent's reply right there keeps the human in charge.
@TSUTSU the problem is personal to me, honestly. We needed better from the review side.