Run several Claude Code agents at once.
Ownlight is a desktop app that gives every Claude Code agent its own live terminal on one canvas. It runs the Claude Code you already have installed, on your own Claude plan or API key, and shows you what each agent asked, did and answered.
Free with an Ownlight account, three agents at once. Pro from $10 a month. Apple silicon Macs, 64-bit Windows 10 or 11 and 64-bit Linux.
Every agent in its own terminal
Open a folder, pick Claude Code in New agents and it starts in a tile on the canvas. Start as many as you need, in one project or several, arrange the tiles and save the layout with the workspace. Each terminal’s header names the model and says in one line what that agent was asked to do.
Codex, opencode, DeepSeek Harness, Grok Build, Gemini CLI and Antigravity run on the same canvas, so you can mix them. See Codex in Ownlight.
Read the answer, not the scrollback
Simple mode shows each turn as what you asked, one line for how it went, and the answer, with paths you can click and a card for each thing the turn made: files, pages, screenshots, running apps and pull requests. It reads Claude Code’s own transcript on your computer, so no model rewrites it and it costs no tokens.
Press Cmd+R, or Ctrl+Shift+R on Windows, to switch every agent back to its terminal. Read how Simple mode shows a turn.
Questions and permissions come to you
When an agent asks a question or wants permission, even from a helper agent it started, the request pins itself above that agent’s feed. Permission prompts get Allow once and Deny. Needs you lists every agent that is waiting on you, and Ownlight can say out loud which one needs you and which one has finished.
Your plan’s limits in the top bar
Small meters show your Claude Code usage for the 5-hour window and the week, in place of a cost line under every agent. For plans billed by usage, such as Enterprise, the meter shows the month’s spend.
Add more than one Claude account and the meter’s panel lists each one with its own usage. New agents start on the account you pick, with the same settings, hooks and history on every account.
Review the branch before you merge
Pro mode puts one agent in the middle with its branch and shows its changes on the right: the whole branch, what isn’t committed, the last turn or one commit. Comment on a line and Claude Code gets the comment. Revert a file or a hunk. Checks and pull requests come from the GitHub CLI you’re signed in to, and you can create, mark ready and merge a pull request from there. After a merge, Ownlight can archive the agent, remove its worktree and pull the base. Pro mode is part of Pro.
Archive when it’s done
Archive closes a finished agent’s terminal and moves it to History with its whole conversation. Resume opens it again where it was, ready for the next prompt. Agents at work, agents that need you and pinned agents are never archived.
Claude Code on a local model
Local can start Claude Code on a model your own machine serves, beside your cloud agents. Read about local models in Ownlight and see how models compare on Ownlight Bench.
Questions
Does Ownlight replace Claude Code?
No. It runs the Claude Code CLI you already have, in a real terminal, on your Claude plan or API key. Ownlight doesn’t include an AI subscription; install the agents you want to use and Ownlight finds them when you open New agents.
How many Claude Code agents can I run at once?
Three on Free. Pro has no limit. Your Claude plan’s own usage limits still apply, and the meters in the top bar show where you are.
Do my prompts go through Ownlight’s servers?
No. Claude Code talks to Anthropic, or the provider you set up, the way it does in your own terminal. Ownlight doesn’t pass your prompts through a server of its own.
Does it run on Windows?
Yes. Ownlight runs on 64-bit Windows 10 and 11 with a signed installer, and on Macs with Apple silicon. Your plan covers both. Intel Macs aren’t supported yet.
Try it on your own agents
Ownlight is free to download and free to use for three agents at once.