Use your ChatGPT subscription as an API

Ways to use a ChatGPT plan from code

A ChatGPT subscription gives you no API key on its own. Below are the four ways to get OpenAI models into your own code.

Plus, Go and Free include no API usage and get no API credits. On Pro, the $100, $200 or $500 a month in API credits must be linked to an API organization you own, expire at the end of every billing cycle, can't pay for Codex usage through API keys and don't raise your signed-in Codex limits. Plan2API is the option that runs on the Codex usage your plan already includes.

Connect ChatGPT in 3 steps

1. Sign up at plan2api.com/connect with a Google account.
2. Connect ChatGPT with a one-time device code: open the link Plan2API shows you, sign in to ChatGPT on OpenAI's own page and enter the code.
3. Copy the p2a_ API key you get after connecting (the dashboard Quickstart page has copy-paste setup for curl, Python and Node).

Here's the official OpenAI Python SDK pointed at https://api.plan2api.com/v1 and calling openai/gpt-6.1-sol.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.plan2api.com/v1",
    api_key="p2a_...",
)

reply = client.chat.completions.create(
    model="openai/gpt-6.1-sol",
    messages=[{"role": "user", "content": "Summarize this diff in one line: ..."}],
)
print(reply.choices[0].message.content)

Codex models and what they cost on the API

The table below lists the Codex models on ChatGPT plans, OpenAI's own estimate of local messages per five hours on Plus, and what the same models cost per million tokens on the pay-per-token OpenAI API.

• Plan2API reads the Codex model list live from OpenAI and exposes each model as openai/<model> (codex/<model> works too), such as openai/gpt-6.1-sol or openai/gpt-6-astra; requests run on your own ChatGPT plan, so pick a model your plan includes.
• GET /v1/models lists every model Plan2API serves, including providers you haven't connected, and a model whose provider isn't connected returns HTTP 403 no_credentials.
• The five-hour message counts are OpenAI's estimates rather than fixed limits, and OpenAI says API token prices shouldn't be used to estimate included subscription usage.

What works and what doesn't

• Requests use the OpenAI Chat Completions format on POST /v1/chat/completions (see the API docs).
• Streaming works.
• Tool and function calling work on Codex models.
• Reasoning effort works via reasoning_effort or OpenRouter-style reasoning.effort.
• Image input works on Codex models that accept images, and image URLs in messages are passed to OpenAI as sent instead of being fetched by Plan2API.
• JSON mode works with response_format set to json_object or json_schema.
• temperature and top_p are ignored on Codex models, and there is no /v1/responses endpoint, so the Codex CLI itself can't point at Plan2API (it already runs on your plan anyway).
• The API sends Access-Control-Allow-Origin: * CORS headers, so browser apps can call it directly alongside scripts and servers.

Usage limits per ChatGPT plan

Plan2API adds no token quota of its own, so your ChatGPT plan's Codex usage limits apply directly.

You can check what you have left in the Codex usage dashboard or by running /status in the Codex CLI, and Plus and Pro subscribers who hit their limit can buy ChatGPT credits to keep going. OpenAI has given Pro 100 and Pro 500 twice their previous Work and Codex allowance, while Plus usage is unchanged (see ChatGPT Plus vs Pro).

If you connect two or more ChatGPT accounts, a request that hits a limit on one account moves to the next before any output is sent. When every connected account is out of quota, the API returns HTTP 429 with code subscription_quota_exhausted and a Retry-After header set to the earliest reset, or 429 subscription_throttled when OpenAI is rate limiting.

Generate images on your ChatGPT plan

On Plan2API Pro and Scale, POST /v1/images/generations and POST /v1/images/edits render images with gpt-image-2 using your connected ChatGPT (Codex) subscription. On Plan2API Free, both endpoints return HTTP 403 with type plan_limit.

You can pass n from 1 to 4, size sets the aspect ratio only (from 1:3 to 3:1, at about 1.57 megapixels per image), output comes back as PNG in b64_json, and edits take up to 5 source images. A render takes about 15 to 45 seconds and up to about 110 seconds at high quality, so set your client timeouts above 2 minutes; OpenAI counts image generations against the same Codex limits as messages, burning through included usage about 3 to 5 times faster than similar turns, and doesn't include them on ChatGPT Free.

Here is a curl request that generates an image through /v1/images/generations.

curl https://api.plan2api.com/v1/images/generations \
  -H "Authorization: Bearer $PLAN2API_KEY" \
  -H "Content-Type: application/json" \
  --max-time 180 \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A ceramic mug on a walnut desk, soft window light",
    "n": 1,
    "size": "1536x1024"
  }' | jq -r '.data[0].b64_json' | base64 --decode > mug.png

Plan2API or Sign in with ChatGPT?

OpenAI's own Sign in with ChatGPT lets eligible Plus and Pro subscribers use their plan for AI requests inside participating apps like Amp Code, Devin, Kilo Code, Notion, Vercel and Warp, as well as open-source tools including OpenCode, OpenClaw, Pi and T3. Usage in those apps counts toward your normal plan limits, and Settings > Usage > App limits lets you cap each app's weekly share.

That only works inside apps that joined OpenAI's program. For your own scripts, internal tools or anything else that speaks Chat Completions, Plan2API gives you a standard base URL (https://api.plan2api.com/v1) and one p2a_ key that can also route to Claude, Gemini and Kimi subscriptions by model prefix.

Connected logins are encrypted with HPKE (X25519) and AES-256-GCM so only your API key can unlock them, the key's secret is never stored, prompts and outputs are never stored, and Plan2API's servers and database run on OVHcloud in Germany (how Plan2API protects logins). Because you sign in on OpenAI's own page, Plan2API never sees your ChatGPT password.

• Free: $0, 1 connected subscription, 2 API keys, 50 chat requests a day, no card required.
• Pro: $12 a month or $99 a year, 3 connected subscriptions, 10 API keys, no daily request cap, image generation included.
• Scale: $29 a month or $249 a year, 25 connected subscriptions, 50 API keys, no daily request cap, image generation included.