All videos

AgentsRoom Compare agents: the same bug given to Claude Code and Codex, one result kept

Filmed in the real app: a demo shop has a checkout bug, and one request goes to Claude Code and Codex, each in a separate copy of the repo, with their panes side by side. One card per agent shows files changed, lines added and removed and working time, and View diff shows everything it did. The app never names a winner: here the smaller change is the one kept, and after a confirmation it is merged into main.

Read the full feature page

Transcript

Same bug, two AI agents. Read both fixes, keep your pick. This is AgentsRoom, the desktop app where your AI coding agents work side by side.

Here, a shop project with a checkout bug. Claude Code or Codex for this one? A benchmark tested someone else's repository, not yours.

So give it to both. From the lightning button, choose Compare on several agents. You type one request.

Each agent works in a separate copy of your repo, your main checkout does not move, and each uses its own quota. Their panes open side by side, and a bar follows both. When both have finished, the bar offers Compare.

One card per agent: files changed, lines added and removed, working time. View diff shows everything it did. Claude Code changed four files and added tests.

Codex changed two files, with a smaller fix. The app never names a winner. You read the code and you decide.

I want the small change, so I keep that one. It asks before merging into main, then offers to delete the other copy. The fix is on main.

Compare agents is in AgentsRoom, under the lightning button. The link is in the description.