Authors
Contents
Find out what your AI bill could drop to
A common question, in different words: is it actually against the rules to share a Claude Max or ChatGPT Pro plan across a few people? Rather than paraphrase, this post quotes the documents and links to them, so you can read the wording yourself.
Why this matters
When a team shares a consumer plan, two things are at stake. The first is the account itself: providers can suspend an account that is used outside its terms, and a suspended account takes its history and its workflow with it. The second is your own position. If you sign a vendor's terms and then operate against them, that is a fact a client or an auditor can find.
None of this requires the provider to be unreasonable. The plans are priced and limited for one person. Using them as a team product is simply not what they are for.
What Anthropic says
Anthropic publishes a legal and compliance page for Claude Code. On the question of plan limits it says:
"Advertised usage limits for Pro and Max plans assume ordinary, individual usage of Claude Code and the Agent SDK."
Source: Anthropic, Claude Code legal and compliance, code.claude.com/docs/en/legal-and-compliance.
The important word is individual. The limits, and the price that goes with them, are set for one person using the tool in a normal way. A team rotating through one Max seat is not ordinary individual usage, however reasonable each person's own use looks.
What OpenAI says
OpenAI's Sign in with ChatGPT terms, dated 29 September 2026, include a list of prohibited uses. Two items on that list are directly about sharing:
"Pooling, transferring, reselling, gifting, or sharing ChatGPT plan usage"
"Using one user's subscription to fulfill another user's requests."
Source: OpenAI, Sign in with ChatGPT Terms, openai.com/policies/sign-in-with-chatgpt-terms.
That second line covers the exact pattern many teams fall into: one person's plan, used to answer requests from several people. It does not matter whether the people are colleagues or whether the use is for work.
What the gateway docs say
The same providers describe how an organization is supposed to do this. Anthropic's Claude Code documentation covers routing the tool through an LLM gateway run by your company. The pattern is described plainly:
"the provider key stays server-side; developers hold gateway credentials instead"
Source: Anthropic, Claude Code LLM gateway docs, code.claude.com/docs/en/llm-gateway.
This is the shape of a compliant team setup. The company holds the provider account. Each developer holds a credential issued by the gateway, not the provider key and not a shared login. Claude Code officially supports this.
What this means for your team
Put the three documents together and the picture is simple.
- Consumer plans (Pro, Max, ChatGPT Pro) are for one person. Sharing their usage is prohibited by OpenAI and outside the assumed use at Anthropic.
- Team use is meant to run on accounts licensed for team use: API accounts and Team or Enterprise plans that the company owns.
- The supported way to give many people access to one company account is a gateway that issues each person their own credential.
Note what is not on the list. There is no supported way to turn a consumer plan into an API, to pool several plans into one, or to resell plan usage. Any product that promises that is asking you to operate against the terms above.
The honest way to set it up
Proxen is built to sit inside these rules, on purpose. Your company opens its own API and Team accounts with Anthropic, OpenAI and Google (we help if you have not done this). Those provider keys stay server-side. Proxen issues one key per person on top, with a model list, a monthly budget and a hard cap. Claude Code, Codex, Cursor and your scripts point at the Proxen address and keep working as before.
Every token is then licensed to your company, every request has a named owner, and nothing in the setup depends on a grey area. If you want to check your current setup against the documents above, talk to us. We will tell you plainly what is fine and what is not.
