login-with-chatgpt 0.1.1

Use the OpenAI Python SDK with your own ChatGPT subscription

login-with-chatgpt

Use the official openai Python package with your own ChatGPT subscription. Authentication follows the OAuth and ChatGPT-backed Codex transport implemented by the reference login-with-chatgpt project.

[!WARNING] This package uses a private, undocumented ChatGPT Codex endpoint. It is not the OpenAI API, has no compatibility or availability guarantee, and may stop working when the upstream protocol changes. Requests consume the signed-in user's own ChatGPT plan.

Install and sign in

uvx login-with-chatgpt login
uv add login-with-chatgpt

Browser PKCE is the default. For SSH, containers, or other headless environments:

uvx login-with-chatgpt login --device

Credentials are stored in the operating system keyring. The SDK never falls back to a plaintext token file.

OpenAI-compatible Responses client

login_with_chatgpt.OpenAI() returns an actual openai.OpenAI instance, so upstream response types, exceptions, parsing, streaming, and tool helpers remain available.

from login_with_chatgpt import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-5.6-sol",
    input="Explain Python descriptors in three sentences.",
)
print(response.output_text)

Async code uses AsyncOpenAI().

Structured output

from pydantic import BaseModel
from login_with_chatgpt import OpenAI


class Answer(BaseModel):
    summary: str
    points: list[str]


client = OpenAI()
response = client.responses.parse(
    model="gpt-5.6-sol",
    input="Summarize OAuth PKCE.",
    text_format=Answer,
)
print(response.output_parsed)

The supported surface is deliberately limited to POST /responses. Use ChatGPTAccount().list_models() or login-with-chatgpt models for model listing.

Guaranteed request features are text, SSE streaming, Pydantic structured output, function tools, image URL/data URL input, function-result continuation, and stateless multi-turn input replay. The SDK does not execute function tools.

Stateful response parameters (previous_response_id, conversation, background mode, and store=True) are rejected. Output token limits are also rejected because the private endpoint does not honor them; the SDK never silently removes a caller-specified usage limit.

Profiles and diagnostics

login-with-chatgpt --profile work login
login-with-chatgpt profiles
login-with-chatgpt use work
login-with-chatgpt status
login-with-chatgpt models
login-with-chatgpt doctor
login-with-chatgpt logout

Profile precedence is: explicit --profile / Python argument, LOGIN_WITH_CHATGPT_PROFILE, active profile, then default.

The Codex protocol version is pinned to the version tested by this package. For an urgent compatibility override, set LOGIN_WITH_CHATGPT_CLIENT_VERSION or pass client_version="0.144.1" to OpenAI(), AsyncOpenAI(), or ChatGPTAccount(). doctor compares the effective protocol version with an installed Codex CLI when one is available; the CLI is not a package dependency.

Development

uv sync --dev
uv run ruff check .
uv run pyright
uv run pytest
uv build

Real-account contract checks are opt-in because they consume subscription usage:

$env:LOGIN_WITH_CHATGPT_LIVE_MODEL = "gpt-5.6-sol"
uv run pytest -m live

On POSIX shells, use export instead of set. Live tests use the currently active credential profile and are never run by CI.