Use spare Codex capacity.
Have a room invitation? Run this command to use shared capacity in your own Codex project.
Requires Node 22.13+ and Codex CLI or desktop app
Who can see my prompts?How the room works
- Using capacity
- Ask the room operator for a single-use invitation. Enter it in the client, name your device, then choose a project. You can use pool credits without a Codex subscription when a seller has capacity. In Auto mode, an eligible personal subscription is used first.
- Sharing capacity
- The operator connects the seller account with a private seller profile. Sellers can offer a percentage manually or let Auto learn from at least 24 hours of usage and reserve capacity for their own work before a reset. An offer may expire unused.
- What a credit buys
- Credits cover model usage inside the room. After a request finishes, 99% of the charge goes to the seller pool and 1% is the service fee. A request already running can exceed the remaining balance. Live payments and cash payouts are disabled.
Who can see my prompts?
- On your computer
- Codex runs tools and opens project files locally. When it includes a code excerpt or tool result in a prompt, that content travels with the request.
- Along the route
- On your own allowance, Auto sends the model request directly to OpenAI. On pool credits, the hosted client sends it over HTTPS through Cloudflare to the room gateway, which forwards it to OpenAI using the seller’s login. The gateway operator can read prompts and responses. Choose a room you trust with the project context you send.
- In the room
- The seller dashboard shows account and device labels and usage totals. Prompts and code stay out of that dashboard. Buyers use separate device credentials and never receive the seller’s login. You can revoke a linked device under Account → Devices.
Sandbox order book
Illustrative Codex market
Last sample trade
— USD / 1M
| USD / 1M | Tokens | Expiry |
|---|---|---|
| Loading buy orders… | ||
| USD / 1M | Tokens | Expiry |
|---|---|---|
| Loading sell orders… | ||
Loading order book…1% fee on filled tokens
Codex API reference
GPT-5.3-Codex · USD per 1M tokens
- Input
- $1.75
- Cached input
- $0.175
- Output
- $14.00