OpenAI is switching off what your AI runs on.

Agent Builder, prompt objects, GPT‑5, GPT‑4o, GPT‑4.1, Whisper, TTS and every realtime voice model have fixed shutdown dates. After that date, calls fail with model_not_found. Scan your code or automation exports for free. If you'd rather not fix it yourself, we migrate and test it before the deadline.

Free shutdown scanner

Paste code, a .env, or an n8n / Make / Zapier export, or drop files or a whole folder. It runs in your browser, so nothing is uploaded or stored.

Drop files or a folder here, or ·
Data: OpenAI deprecations page, checked .

Want it fixed before the deadline?

Fixed prices, delivered in 24–48 hours. Reserve a slot now; you get a written scope and payment details before any work starts. We never need your passwords: you invite us to a repo or send an export, and you run the test script with your own API key.

Shutdown Audit

$199
24 hours · credited if you upgrade
  • Every retiring model, workflow and prompt object, with file and line
  • Risk ranking: what breaks first and what changes behaviour
  • Replacement and cost difference per call
  • Step-by-step fix plan
Get the audit

Model Migration

$549
48 hours · 1 project
  • Up to 10 AI calls or workflows moved to the recommended replacements
  • Parameter fixes (reasoning models, token limits, JSON output)
  • Side-by-side regression report on 20 real test cases
  • 7 days of follow-up fixes
Book a migration

Agent Builder Rescue

$990
Live before Nov 30 · 1 workflow
  • Your Agent Builder workflow rebuilt on the Agents SDK, same behaviour
  • Tools, guardrails and handoffs carried over and tested
  • Your ChatKit widget pointed at the new backend
  • Deployed to your hosting, with a runbook
Rescue my agent

Agency Pack

$1,490
Up to 5 client projects · white-label
  • Audit + migration for 5 client builds
  • Reports under your brand
  • 50% deposit books one of the pre‑Dec‑11 slots
Book agency slots

Guarantee: if a migration isn't delivered on time, or doesn't pass the agreed test cases, you get a full refund.

How it works

1 · Scan or sendRun the scanner, or send us a repo invite or export files.
2 · Fixed quoteYou get a fixed price and a delivery time before paying.
3 · Migrate and prove itWe change the code, run old vs new on your test cases and show the differences.
4 · You approveMerge the pull request or import the files. Done before the shutdown.

FAQ

Isn't it just changing one model name?

Sometimes. But the replacements are different models: reasoning settings, token limits, output style, tool calling and cost can all change, so prompts that worked can quietly get worse. Agent Builder is bigger still: OpenAI's export does not convert the workflow graph or guarantee identical behaviour. What you're paying for is the test report that shows nothing broke.

Who are you?

SunsetFix is run by Riccardo, an independent developer in Rome, Italy. The work is done with AI coding tools (Claude), and every change is tested against your cases and explained before you merge it.

Do you need my API key or passwords?

No. You keep your keys. We send a test script that you run with your own key, or you add a temporary key to a test environment you control.

Which dates matter?

Agent Builder, the Evals API and reusable prompt objects: Nov 30, 2026. GPT‑5 / GPT‑5 mini / GPT‑5 nano / o3: Dec 11, 2026. TTS: Jan 6, 2027. GPT‑4o / GPT‑4o mini / GPT‑4.1 / GPT‑4.1 mini: Jan 14, 2027. Realtime and audio models: Jan 20, 2027. Whisper and transcribe models: Feb 26, 2027. Always check OpenAI's page for your exact model IDs.

Is this affiliated with OpenAI?

No. SunsetFix is independent and not affiliated with or endorsed by OpenAI.