Billing, terms and who can use your agent
An agent runs on your own plan or key and answers only you. Here is why Omne works this way, what it keeps, and what the makers' terms say.
Your agent is billed by its maker, on your own plan or key, and never from your Omne credits. Only you can use it, and Omne never holds your sign-in with the maker.
Billing
The panel for each agent has a Billed row. For Claude Code it reads your Claude plan or your Anthropic key, for Codex your ChatGPT plan or your OpenAI key, and so on, followed by "Never your Omne credits." The Connected list also shows how each agent runs, as reported by your computer: On your own plan, On your own API key, or On your own account when it cannot tell.
Who can use it
Only the person who connected it. Teammates do not see your agents and cannot reach them. Omne checks this on the server for every turn, not only on screen. The reason is in the makers' consumer terms: an account on a personal plan cannot be made available to anyone else. Omne does not offer a way around that, such as sharing an agent with a team on a personal plan.
What Omne keeps
Omne keeps the agent's name, the machine name your computer reported, how it signs in, and a key for the connection. Your sign-in with the maker stays on your computer. Omne never sees it, never routes requests through it, and never resells its use.
Why it works this way
Anthropic's published rules allow a person to sign in to its unmodified Claude Code program with their own subscription, even when another tool starts that program. They do not allow another app to collect or hold the login, or to route requests through the plan for users. Omne's connector starts the maker's own program and holds nothing. OpenAI documents API keys as the recommended way to automate and documents device sign-in for machines without a browser. This is a reading of public terms, not legal advice, and a business should check the terms that apply to it. Anthropic's consumer terms also carry a non-commercial use line, so a business on a consumer plan may want to confirm it is allowed.
Limits on the makers' side assume ordinary, individual use, which is why Omne's tools work only during a turn you started. See What an agent can do in Omne.
When your computer is off
An agent runs on your computer, so it is offline when the computer is. Each agent's panel has a section named Need it when your computer is off? with a shortcut to the same maker's API key or plan under AI providers. For Claude Code that is your Anthropic API key, for Codex your ChatGPT plan, for Gemini CLI your Gemini key and for Grok Build your Grok plan. Note that the ChatGPT and Grok plan sign-ins are still coming soon, as described in Use your ChatGPT or Grok plan. Running the maker's own program on Omne's servers is planned, not built.
Costs you can still see
Each maker has its own usage page. The cost of a chat is on your plan's allowance or your key's balance with that maker. Omne's AI usage page does not read an agent's use.
Last updated November 3, 2026
There's a real person behind Omne, so ask anything.