$59 first monthPlans
unlimitedcodex
Packages
Page language

Codex CLI / IDE setup guide

Connect your unlimitedcodex key to Codex safely.

Run Codex with your unlimitedcodex key

This page helps you run Codex CLI, Codex IDE, or the Codex app with your unlimitedcodex API key and base URL. If you are new, choose one path, enter the key, restart Codex, then run one tiny test.

The current Codex config uses the Responses API: wire_api = responses, requires_openai_auth = true, the recommended model is gpt-5.5, and base_url does not include /v1.
Windows auto setupmacOS / Linux setupTXT for local AI assistant

Step 1

Choose the setup path you trust most.

You do not have to run a script. Use Windows auto setup, macOS/Linux command setup, or manual setup. You can also give the TXT instructions to a trusted local AI assistant.

Windows automatic

The BAT file backs up config.toml and auth.json, asks for your key, and writes the settings into your local .codex folder.

Download BAT

macOS / Linux

Run the command file in Terminal. iPhone and iPad cannot perform this local Codex setup.

Download COMMAND

Controlled AI help

Download the TXT file and give it to a trusted local/private AI assistant. The assistant should only edit local config files.

Download TXT

Checklist

Check this before you start

Most setup problems come from the wrong folder, a /v1 suffix in base_url, or forgetting to restart Codex.

1

Do you have the API key?

Your key arrives in the delivery email. Put it only in auth.json or where the setup file asks for it.

2

Has Codex opened once?

If config.toml or auth.json is missing, sign in to ChatGPT inside Codex, close and reopen Codex, then check again.

3

Is the base URL correct?

In config.toml, base_url should be only https://claude.kadirr.dev. Do not put /v1 there.

4

Restart is required

After editing config.toml and auth.json, fully close and reopen Codex CLI, Codex IDE, or the Codex app.

New model selection

Choose the model first, then copy the config.

The active model for new setup is gpt-5.5. Keep the model line on gpt-5.5 while 5.6 Sol and its variants stabilize.

5.6 models are temporarily paused

5.6 Sol and its variants are temporarily restricted because of upstream instability and slowness. Use gpt-5.5 for all new setups until support says otherwise.

Reasoning effort

xhigh is the safe default for coding and agent work. Choose high, medium, or low for faster and simpler jobs.

Manual setup

Edit these two files if you do not want to run a setup file.

config.toml tells Codex which API provider to use. auth.json stores your unlimitedcodex API key locally.

Windows folder: %USERPROFILE%\.codex
macOS / Linux folder: ~/.codex

If the files are missing, that is normal

With newer Codex versions, config.toml or auth.json may appear only after you sign in to ChatGPT inside Codex and open Codex once.

Paste into config.toml

model_provider = "OpenAI"
model = "gpt-5.5"
review_model = "gpt-5.5"
model_reasoning_effort = "xhigh"
disable_response_storage = true
network_access = "enabled"

[model_providers.OpenAI]
name = "OpenAI"
base_url = "https://claude.kadirr.dev"
wire_api = "responses"
requires_openai_auth = true

[features]
goals = true

Paste into auth.json

{
  "OPENAI_API_KEY": "YOUR_UNLIMITEDCODEX_API_KEY"
}

Small API-level test

After restarting Codex, run one short test before a large repository task. If it fails, check the key, package status, and base URL.

curl -X POST https://claude.kadirr.dev/v1/responses \
  -H "Authorization: Bearer YOUR_UNLIMITEDCODEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "input": [
      {
        "role": "user",
        "content": [
          {
            "type": "input_text",
            "text": "Hello from unlimitedcodex"
          }
        ]
      }
    ],
    "stream": false
  }'

Debug

Common problems

Check this table before asking on Telegram. Do not send your full API key to support.

401 or invalid_api_key

The key may be wrong, inactive, truncated, or written into the wrong file.

Check only the OPENAI_API_KEY field in auth.json. Do not paste the full key into public places.

404 or route not found

This usually happens when /v1 is added to base_url inside config.toml.

Keep base_url as https://claude.kadirr.dev. Use /v1/responses only in direct API tests.

Settings look unchanged

Codex may still be running with the old process.

Fully close and reopen Codex CLI, Codex IDE, or the Codex app.

config.toml or auth.json is missing

Codex may not have created local files yet.

Sign in to ChatGPT inside Codex, open Codex once, then check the .codex folder again.

Need help?

Message us on Telegram. Do not send the full key. Send your OS, error text, which file you edited, and only a safe key prefix.

Telegram support
unlimitedcodex Codex CLI / IDE Setup Guide | unlimitedcodex