Codex CLI Custom API Setup: Base URL and API Key

Before you begin

1. Install Codex CLI using the instructions on the official Codex CLI page.
2. Create your codexinfo.top account and generate an API key. Existing users can use the Sign in button in the navigation.
3. Choose an exact model name from the current model and pricing list.

Step 1: Set the API key environment variable

macOS or Linux

export CODEXINFO_API_KEY="sk-your-api-key"

For a persistent Zsh setting:

echo 'export CODEXINFO_API_KEY="sk-your-api-key"' >> ~/.zshrc
source ~/.zshrc

Use ~/.bashrc instead when your shell is Bash.

Windows PowerShell

$env:CODEXINFO_API_KEY="sk-your-api-key"

To save it for the current Windows user:

[Environment]::SetEnvironmentVariable(
  "CODEXINFO_API_KEY",
  "sk-your-api-key",
  "User"
)

Close and reopen PowerShell after saving a user environment variable.

Step 2: Edit the user-level config.toml

model = "gpt-5.5"
model_provider = "codexinfo"

[model_providers.codexinfo]
name = "codexinfo.top"
base_url = "https://api.codexinfo.top/v1"
env_key = "CODEXINFO_API_KEY"
wire_api = "responses"

gpt-5.5 is an example visible in the codexinfo.top model list when this guide was published. Replace it with the exact current model name when the list changes.

Step 3: Verify the connection

codex --version

mkdir codex-api-test
cd codex-api-test
codex

You can also make a non-interactive test request:

codex exec "Reply with exactly: connection successful"

After a successful request, check the codexinfo.top dashboard for the corresponding usage record.

Troubleshooting

401 Unauthorized or Invalid token

• Confirm that env_key exactly matches CODEXINFO_API_KEY.
• Run echo $CODEXINFO_API_KEY on macOS/Linux or $env:CODEXINFO_API_KEY in PowerShell.
• Reopen the terminal after saving a persistent variable.
• Copy the key again without spaces or line breaks.

404 Responses endpoint not found

• Use https://api.codexinfo.top/v1 as the base URL.
• Do not use the dashboard URL as the API endpoint.
• Make sure the provider supports the Responses API.

Model not found

• Model names must match the dashboard exactly.
• Do not copy an outdated model name from an old tutorial.
• Review the current model list.

429 Too many requests or insufficient quota

• Check account balance and API-key quota.
• Review model, IP, and request restrictions on the key.
• Reduce concurrency and retry later.

Codex ignores the provider configuration

• Confirm that you edited the user-level ~/.codex/config.toml.
• Check TOML quotes, spelling, and table names.
• Restart the Codex session and terminal.
• Do not place the provider definition only in a project-level config file.

API key safety

• Never commit a key to Git.
• Do not expose a complete key in screenshots or support logs.
• Create separate keys for different devices or projects.
• Revoke and replace a key immediately if it leaks.
• Use quota, model, and IP restrictions when appropriate.

Frequently asked questions

Official references

The fields in this guide were checked against the OpenAI Codex Configuration Reference and Codex Basic Configuration. Available models and service pricing come from the live codexinfo.top dashboard.