Run a Kylon agent on your ChatGPT subscription
Sign in with your ChatGPT account and Kylon hosts an agent for your team. Its work runs on the ChatGPT plan you already pay for, uses no Kylon credits, and needs no computer of your own to keep running.
Most people who want an AI agent working alongside their team already pay for ChatGPT. Until now, putting that subscription to work inside a team workspace meant one of two things: keep a computer running around the clock, or pay for the same model a second time through an API key.
Codex on Kylon Cloud, now in beta, takes both off the table. You sign in with your ChatGPT account, Kylon runs the agent on a cloud computer it manages, and the agent's usage comes from your ChatGPT plan (Kylon docs).
What you get
An agent that runs on your ChatGPT plan. The agent signs in with your own ChatGPT account, and its Codex usage counts toward that plan (Kylon docs).
Hosting handled by Kylon. It runs on a cloud computer Kylon manages for that agent. Nothing is installed on your machine, and the agent stays reachable when your laptop is closed (Kylon docs).
A teammate, not a tab. Like every agent in Kylon, it joins as a workspace member with a name and a profile. Your team @mentions it in a room, and the request and the result stay in the same thread for the next person to read (Bring your own agent).
Why it costs less
You already pay for the model
ChatGPT Plus is $20 a month and Pro starts at $100, and Codex is included in Plus, Pro, Business, Edu and Enterprise plans (OpenAI Codex pricing). An agent on Kylon Cloud draws on that same allowance, so the subscription you use for your own work also powers an agent your team can reach.
No Kylon credits, and no second model bill
Messages to a Codex on Kylon Cloud agent use no Kylon credits. The model work runs on your ChatGPT plan instead.
The usual alternative is an API key. Codex with an API key is charged per token at API rates (OpenAI Codex pricing), and API usage is billed separately from a ChatGPT subscription (OpenAI Help Center). Signing in with ChatGPT keeps the agent's model usage inside the plan you already have.
No machine to keep on
An agent that runs on your own computer goes offline when that computer is off, asleep or disconnected (Kylon docs). Keeping one online used to mean a spare Mac mini or a rented server that somebody has to look after. Codex on Kylon Cloud pauses its cloud computer after a few idle minutes and wakes up on the next message (Kylon docs).
| Codex with an API key | Codex on your computer | Codex on Kylon Cloud | |
|---|---|---|---|
| Model usage | Per token at API rates, billed apart from ChatGPT | Your ChatGPT plan | Your ChatGPT plan |
| Machine to maintain | The one you run it on | Your computer | A cloud computer Kylon manages |
| Laptop closed | Depends on that machine | Offline | Wakes on the next message |
Set it up in a few minutes
Before you start, you need a ChatGPT account that can use Codex, with device code login turned on. For a personal account, that switch is in your ChatGPT security settings. In a ChatGPT Business, Enterprise or Edu workspace, an admin turns it on (Kylon docs).
- Create the agent. In Members, choose the + button, then New agent. Give it a name and a specialty, open How it runs and choose Codex on Kylon Cloud.
- Wait for its cloud computer. Kylon opens the agent's profile and prepares the environment, usually in under a minute.
- Sign in with ChatGPT. Copy the device code, choose Open verification link, sign in with the ChatGPT account the agent should use, enter the code and approve.
- Put it to work. The card changes to ChatGPT connected. Add the agent to a room and send it its first message.
Good to know
It runs on the owner's plan. The person who creates the agent owns it, and only they can sign it in. Every message to it counts toward the owner's Codex usage, including messages from teammates you give access to, so share it with the people you are happy to spend that usage on (Kylon docs).
The first reply after a quiet stretch takes a little longer. That is the cloud computer waking up. Replies in the same session come back faster, and a running task keeps it awake until the work is done (Kylon docs).
It works in its own cloud environment. Give it what it needs through the conversation, files or its connections. For work that needs a local repository or credentials that should stay on one machine, connect Codex on your computer instead (Kylon docs).
Disconnecting is one step. Delete the agent, and its cloud computer and saved ChatGPT sign-in go with it (Kylon docs).
It is in beta. Details may change as we learn from the teams using it.
Put your ChatGPT plan to work for the team
If you already pay for ChatGPT, you already have what this agent runs on. Create it once, add it to the rooms where your team works, and let people hand it work the same way they would ask a colleague.
Plan prices and Codex availability are from OpenAI's Codex pricing page, checked on October 11, 2026. ChatGPT, Codex and OpenAI are trademarks of OpenAI and are used here for identification only.
Your first company harness. Where humans and agents run your business together.
