Hi, we're the team behind Orcyx. Running one coding agent is easy. Running three on the same repo is where it gets uncomfortable: they edit the same files, they share one working directory, and you end up as the person untangling it. We saw exactly that in a timed first run. Three agents, one task, and all three wrote to the same path. The second merge stopped on a conflict before anything reached main, which is what it should do, and a good reminder of why they shouldn't share a folder in the first place. So Orcyx gives each agent its own git worktree and its own terminal pane, side by side in one window. When an agent finishes, an auditor reads what it did and flags problems. The merge is still your decision, and a conflict shows up before it touches main. Around that sits the rest of the working day: an IDE tied to the terminals, a browser pane, voice dictation, scheduled automations, and a markdown knowledge graph on your own disk that agents can query through an MCP server. It runs locally. You use your own Claude Code, Codex and other CLI accounts and provider keys, and Orcyx adds no markup to them. Where it stands: pre-1.0 and Windows only for now, with macOS and Linux coming. There is no free tier or trial, and every plan has a seven-day money-back guarantee. The videos are recorded in the real app with real agents. The captions, cursor and title cards are added in editing. How do you keep several agents from stepping on each other today? We're curious where it breaks for you.

Orcyx

Run coding agents side by side, each in its own worktree

Run coding agents side by side, each in its own worktree