Codex setup

This guide gets Codex connected to Modelflare and running its first request. Use Option A for automated setup. Use Option B to install Codex first, then create each required file yourself.

1. Create an API key

Sign in to the Modelflare console, open API Keys, and create a key that can access gpt-5.6-sol. Choose any available non-default group that lists the model; there is no single required group.

Keep the key ready. You can use the installer or configure the same files manually.

2. Choose a setup method

Option A: run the installer

Not comfortable with the terminal yet? Start with the interactive first-run guide. It is only a visual rehearsal and does not change your computer.

macOS or Linux

Run this command as your normal user:

bash

curl -fsSL https://modelflare.dev/install/codex.sh | sh

When you are already logged in as root on a VPS or container, use:

bash

curl -fsSL https://modelflare.dev/install/codex.sh | MODELFLARE_ALLOW_ROOT=1 sh

Do not add sudo on a normal personal computer. The script installs or updates Codex, backs up existing files, asks for your Modelflare API key, and writes the provider configuration under ~/.codex.

Windows PowerShell

Open PowerShell as your normal Windows user and run:

PowerShell

irm https://modelflare.dev/install/codex.ps1 | iex

The installer asks for a language and your Modelflare API key, verifies access to gpt-5.6-sol, and backs up existing files before changing them. When it finishes, continue to Verify the connection.

Option B: create each Codex file yourself

Use this option when you do not want to run the Modelflare setup script. Follow the steps in order and use the command block for your operating system in every step.

Option B creates two files under your user-level .codex directory: config.toml and auth.json.

Step 1: install Codex CLI

Windows PowerShell

PowerShell

irm https://chatgpt.com/codex/install.ps1 | iex
codex --version

Linux

bash

curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex --version

macOS

bash

curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex --version

Step 2: create and enter the .codex directory

Windows PowerShell

PowerShell

New-Item -ItemType Directory -Force "$HOME\.codex" | Out-Null
Set-Location "$HOME\.codex"

Linux

bash

mkdir -p ~/.codex
cd ~/.codex

macOS

bash

mkdir -p ~/.codex
cd ~/.codex

Step 3: create config.toml

Windows PowerShell

PowerShell

$config = @'
cli_auth_credentials_store = "file"
model_provider = "modelflare"
model = "gpt-5.6-sol"
model_reasoning_effort = "xhigh"
disable_response_storage = true
model_supports_reasoning_summaries = true

[model_providers.modelflare]
name = "Modelflare"
base_url = "https://modelflare.dev/v1"
wire_api = "responses"
requires_openai_auth = true
'@
$utf8NoBom = New-Object System.Text.UTF8Encoding -ArgumentList $false
[System.IO.File]::WriteAllText(
  (Join-Path (Get-Location) "config.toml"),
  $config,
  $utf8NoBom
)

Linux

bash

cat > config.toml << 'EOF'
cli_auth_credentials_store = "file"
model_provider = "modelflare"
model = "gpt-5.6-sol"
model_reasoning_effort = "xhigh"
disable_response_storage = true
model_supports_reasoning_summaries = true

[model_providers.modelflare]
name = "Modelflare"
base_url = "https://modelflare.dev/v1"
wire_api = "responses"
requires_openai_auth = true
EOF
chmod 600 config.toml

macOS

bash

cat > config.toml << 'EOF'
cli_auth_credentials_store = "file"
model_provider = "modelflare"
model = "gpt-5.6-sol"
model_reasoning_effort = "xhigh"
disable_response_storage = true
model_supports_reasoning_summaries = true

[model_providers.modelflare]
name = "Modelflare"
base_url = "https://modelflare.dev/v1"
wire_api = "responses"
requires_openai_auth = true
EOF
chmod 600 config.toml

Step 4: create auth.json

Replace YOUR_MODELFLARE_API_KEY in the JSON object with the complete API key you created in the Modelflare console. This file must remain valid JSON and use the OPENAI_API_KEY field; do not add a provider auth command block. Current Codex releases require requires_openai_auth = true for this API-key path.

Windows PowerShell

PowerShell

$authJson = @'
{
  "auth_mode": "apikey",
  "OPENAI_API_KEY": "YOUR_MODELFLARE_API_KEY"
}
'@
$utf8NoBom = New-Object System.Text.UTF8Encoding -ArgumentList $false
[System.IO.File]::WriteAllText(
  (Join-Path (Get-Location) "auth.json"),
  $authJson,
  $utf8NoBom
)

Linux

bash

cat > auth.json << 'EOF'
{
  "auth_mode": "apikey",
  "OPENAI_API_KEY": "YOUR_MODELFLARE_API_KEY"
}
EOF
chmod 600 auth.json

macOS

bash

cat > auth.json << 'EOF'
{
  "auth_mode": "apikey",
  "OPENAI_API_KEY": "YOUR_MODELFLARE_API_KEY"
}
EOF
chmod 600 auth.json

Step 5: start Codex

Windows PowerShell

PowerShell

Set-Location $HOME
New-Item -ItemType Directory -Force "my-codex-project" | Out-Null
Set-Location "my-codex-project"
codex --model gpt-5.6-sol

Linux

bash

cd ~
mkdir -p my-codex-project
cd my-codex-project
codex --model gpt-5.6-sol

macOS

bash

cd ~
mkdir -p my-codex-project
cd my-codex-project
codex --model gpt-5.6-sol

3. Verify the connection

If you used the installer, fully close any Codex CLI, IDE, or desktop session that was already open. Open a new terminal, enter a project directory, and run:

bash

codex --model gpt-5.6-sol

After finishing setup, you can also download the Codex App.

Then ask Codex:

Text

Reply with one sentence confirming the active model and provider.

A normal response means setup is complete.

Common problems

Symptom	What to do
401	Recopy the API key and confirm that it has not been deleted or disabled.
model_not_found	Edit the API key and choose a group that can access gpt-5.6-sol.
ChatGPT login appears	Fully quit every Codex session and reopen the CLI.
codex is not found	Open a new terminal. On Windows, try codex.cmd.