> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aixy-gateway.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Codex with a ChatGPT subscription

> Use each person's ChatGPT plan in the official Codex client through Aixy.

Connect the official Codex client to Aixy while each person keeps their own ChatGPT sign-in.
The person's Aixy key selects the project and user; the ChatGPT credential authorizes upstream
subscription usage. These identities remain separate.

## Enable and verify

1. Select the project, or **All projects** for an organization default.
2. Open **Providers → Add provider → Codex (ChatGPT subscription)**.
3. Select **Enable connection**. Follow the dashboard **Setup guide** to prepare a personal Aixy
   key and configure the official Codex client.
4. Sign in to the intended ChatGPT workspace in Codex, list its models, and send a short prompt.
5. Open [Activity](/observe/activity) and verify the expected project, user, native model, usage,
   and **Subscription** billing label. Test a full tool call and result exchange before adoption.

Use a compatible Codex version and configure HTTP streaming as shown in the dashboard. The client
retains responsibility for ChatGPT sign-in and token renewal. Aixy stores no ChatGPT access or
refresh token in the provider connection. The gateway sees the access token in transit and
forwards it only to the fixed OpenAI Codex backend.

## Models and usage

Codex receives its native catalog with the user's ChatGPT identity. Use those model IDs without
an Aixy provider prefix. Aixy filters models blocked by the effective organization or project
policy while preserving the remaining entries' native metadata. There is no shared catalog cache
across users, and this catalog does not populate **Models** or **Playground**. A successful catalog
response does not prove a complete model or tool request.

ChatGPT plan limits, workspace permissions, model availability, and credits still apply. A
subscription is not unlimited usage, and Aixy does not pool subscriptions or replace unavailable
capacity with another person's account. An expired upstream login remains an upstream
authentication failure.

Validate [project and user rate limits](/costs/rate-limits) with the chosen Codex version before
rollout. Token-per-minute limits require bounded text input and an explicit output-token limit.
Requests that the client cannot bound are rejected; subscription activation does not bypass these
controls.

Reported subscription tokens contribute to usage totals. Subscription fees and credits are
separate from API spend: Activity labels the billing channel and does not assign an API token
price, a zero-dollar price, or savings to these calls. Mixed sessions show API cost coverage only
for their API calls.

## Compatibility boundary

The connection preserves native Codex request history, tool identifiers, opaque reasoning state,
and incremental server-sent events. It supports the official client's model requests using HTTP
streaming. WebSocket transport, voice, sign-in, plugins, and MCP connections are outside this
model route. Configure [Agent tools and MCP](/agent-tools/overview) separately.

The subscription connection uses a fixed upstream destination. It does not offer cross-provider
translation, routing aliases, or fallback. It does not turn a ChatGPT subscription into a general
API credential for other applications. See [OpenAI's gateway documentation](https://learn.chatgpt.com/docs/enterprise/sign-in-with-chatgpt-through-a-gateway)
for the official client contract and the **API Reference** for Aixy's supported HTTP contract.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.