Feature Request: Allow ChatGPT Subscribers to Use Basic Chat in Third-Party and Self-Hosted Clients
Summary
I’d like OpenAI to allow ChatGPT subscribers to use the standard conversational ChatGPT experience included in their subscription through third-party, open-source, and self-hosted clients.
Basic conversations should use the same ChatGPT conversational allowance available through the official web and mobile applications, rather than consuming the separate Work/Codex usage allowance.
The Problem
I’m a paying ChatGPT Plus subscriber, and I want to use ChatGPT through my own preferred interface, such as Hermes Agent.
Today, I can have ordinary conversations in ChatGPT’s web or mobile applications without consuming my Work/Codex allowance.
However, when I authenticate a third-party client using my ChatGPT subscription, even simple conversations consume my limited Work/Codex allowance.
This means I’m effectively penalized for choosing a different frontend to access a service I’m already paying for.
I don’t want to be restricted to the official web or mobile applications just to enjoy the standard conversational benefits of my subscription.
Proposed Solution
Extend the existing Sign in with ChatGPT functionality to support standard conversational usage through authorized third-party clients.
Specifically:
-
Standard ChatGPT conversations: Allow basic conversations through third-party clients to consume the same conversational allowance available in the official ChatGPT applications, subject to the same applicable limits.
-
Separate usage allowances: Keep basic conversational usage independent of the Work/Codex allowance. Using a terminal, desktop application, or self-hosted frontend shouldn’t automatically classify a conversation as a coding or agentic workload.
-
Existing subscription authentication: Use the Sign in with ChatGPT OAuth integration to authenticate users and authorize access to their existing conversational subscription benefits.
-
Consistent subscription experience: Provide the same eligible models and applicable usage limits for basic conversations, regardless of the client being used.
Example Use Case
I use Hermes Agent as a custom AI interface and would like the flexibility to interact with ChatGPT through it.
My expected usage would be:
| Activity | Expected Allowance |
|---|---|
| General conversations and questions | Standard ChatGPT |
| Technical discussions and troubleshooting | Standard ChatGPT |
| DevOps architecture and planning | Standard ChatGPT |
| Codex CLI coding tasks | Work/Codex |
| Autonomous coding and agentic workloads | Work/Codex |
I fully expect Codex CLI and other agentic coding tools to consume my coding allowance. That’s what it’s intended for.
What I don’t expect is that simply having a normal conversation from my own client should consume that same limited allowance.
What I’m Not Asking For
This is not a request for unrestricted or unlimited API access, additional coding usage, or the removal of existing rate limits and safeguards.
I’m asking for the ability to access the standard conversational features already included in my paid ChatGPT subscription, with the same applicable usage policies, through an authenticated client of my choosing.
Why This Matters
OpenAI has already made progress toward third-party interoperability through Sign in with ChatGPT.
Allowing subscribers to access their standard conversational benefits through external clients would be a meaningful next step.
Users should be able to choose how they interact with ChatGPT, whether through a browser, mobile application, terminal, or self-hosted interface, without sacrificing an entirely separate usage allowance.
I subscribe to ChatGPT for the service and models, not because I specifically want to use a particular frontend.
Requested Outcome
Please introduce a supported way for authorized third-party clients to access the standard ChatGPT conversational allowance independently of Work/Codex usage.
The fundamental request is simple: Let paying subscribers use their included basic ChatGPT conversations through the client of their choice, while keeping coding and agentic usage separate.