Your hardware. Your agents.

View on GitHub

Meet Jelly: open-source agents on your hardware, with your existing ChatGPT subscription. Real mobile UI, private browser handoffs, and sudo—in 29.5 seconds.

Real mobile UI with demonstration data, animated characters, and an original soundtrack. Sudo execution and website sign-in are simulated; no real credentials are entered. Using Jelly requires a ChatGPT subscription or OpenAI API key. Model requests go to OpenAI; your plan’s usage limits apply.

Launch reel transcript
  1. 0:00 — Agents on your hardware. Open source. MIT licensed and self-hosted. Jelly’s mobile agent list appears.
  2. 0:04 — Meet the crew. Laptops out. Jelly plans your day, System runs commands, and Browser drives a real browser. Chat, files, code, commands, and a browser on your own computer.
  3. 0:08 — Ask for anything. It gets to work. Real mobile screen-recording frames show a message, the working state, and an answer.
  4. 0:12 — Need admin rights? Type it privately. System ducks behind the phone: “Not looking!” A private sudo prompt is filled and the simulated command completes. Private authentication is required only when your OS asks; the password is never sent to the agent.
  5. 0:17 — It browses for you. Sign in? Your turn. Browser ducks away: “Still not looking!” Agent browser tools pause during private sign-in. Hand control back and the agent carries on.
  6. 0:22.5 — Bring your ChatGPT subscription. The crew appears above the account settings. Model requests go to OpenAI and plan limits apply.
  7. 0:25.5 — Jelly. Your hardware. Your agents. Open source, MIT licensed, self-hosted. Get Jelly at github.com/dctanner/jelly.