1
Confirm custom base URL support
Before paying for any provider, make sure your Codex IDE, Codex CLI, SDK, or agent client can use a custom OpenAI-compatible base URL instead of a fixed first-party endpoint.
Codex custom base URL guide
Use an OpenAI-compatible API key and base URL in Codex workflows.
This guide shows how to verify a custom provider before a long Codex IDE, Codex CLI, SDK, or agent run: check the base URL, model list, one chat request, optional endpoints, and concurrency fit.
Last updated: July 8, 2026. unlimitedcodex is independent and not affiliated with OpenAI.
Direct answer
What must work before you trust a custom base URL?
A useful provider test proves five things: the client can override the base URL, the API key authenticates, /v1/models returns expected model IDs, one small chat/completions request works, and required endpoints such as embeddings or image generation via API are checked separately.
| Base URL | https://unlimitedcodex.com/v1 Use the delivered OpenAI-compatible base URL after checkout or during an approved test. |
|---|---|
| API key | UCX_API_KEY or client-specific secret Keep the full key in environment variables. Do not paste it into public clients or support threads. |
| Model ID | gpt-5.5 or codex-5.5-style values Use the exact model IDs included in setup delivery or test instructions. |
| Concurrency | 4 concurrent connections Plan agent workers, retries, and background jobs around this package limit. |
| Pre-buy test | https://t.me/unlimitedcodex_com Request a limited 1-hour test when you need proof before paying. |
One-hour verification flow
Keep the first test small enough to diagnose.
The point of a pre-buy test is not to run a full production workload. It is to prove endpoint shape, auth, model IDs, and client routing before money and time go into the wrong configuration.
2
Request a small test first
For unlimitedcodex, serious buyers can request a limited 1-hour Telegram API test before paying. Share the client and endpoint shape you need, but do not send private keys, auth headers, private logs, or secrets.
3
Verify model discovery
Run GET /v1/models or the equivalent model-list command. Confirm the model IDs your workflow expects, such as GPT-5.5 or Codex-style model values, before a long agent run.
4
Send one tiny chat request
Run one minimal chat/completions request with a harmless prompt. This proves authentication, JSON shape, model ID, routing, and response format before larger work begins.
5
Test optional endpoints separately
If your app needs embeddings or image generation via API where supported, test those endpoints independently. Do not assume chat success proves every endpoint is available.
6
Move production config into env vars
After checkout and manual delivery, store the delivered API key and base URL in environment variables, keep secrets out of source control, and plan around 4 concurrent connections.
Smoke tests
Run these before a long Codex agent session.
Use harmless inputs first. If these fail, do not start a long repo edit, migration, or agent loop until the route, key, and model ID are fixed.
Check /models
curl "https://unlimitedcodex.com/v1/models" \ -H "Authorization: Bearer $UCX_API_KEY"
Send a tiny chat request
curl "https://unlimitedcodex.com/v1/chat/completions" \
-H "Authorization: Bearer $UCX_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.5",
"messages": [
{
"role": "user",
"content": "Reply with one short sentence."
}
]
}'Store config safely
UCX_API_KEY="ucx_live_..." OPENAI_API_KEY="$UCX_API_KEY" OPENAI_BASE_URL="https://unlimitedcodex.com/v1"
FAQ
Common setup questions.
Can Codex IDE or Codex CLI use an OpenAI-compatible base URL?
Use this guide only for clients or workflows that support a custom OpenAI-compatible base URL and Bearer API key. Before paying, verify your exact Codex IDE, Codex CLI, SDK, or agent client can route requests to a custom provider.
What should I test before paying for unlimitedcodex?
Test /v1/models, one tiny chat/completions request, model ID handling, route verification, and any endpoint your workflow needs such as embeddings or image generation via API where supported. Also check that a 4 concurrent connection limit fits your workload.
Does unlimitedcodex provide instant self-serve API keys?
No. unlimitedcodex uses manual setup delivery after checkout, usually 10 minutes to 5 hours. Visitors can request a limited 1-hour Telegram API test before paying when they need proof of endpoint fit.
Is unlimitedcodex affiliated with OpenAI?
No. unlimitedcodex is an independent OpenAI-compatible GPT-5.5 and Codex API access provider and is not affiliated with OpenAI.
Next step
Test route fit first, then choose Weekly or Monthly access.
unlimitedcodex packages are $19/week or $76/month, with unlimited token consumption, 4 concurrent connections, and manual setup delivery in 10 minutes to 5 hours after checkout.