Codex Quick Start
Install Codex on Windows, macOS, or Linux, connect a Modelflare API key, and verify a Responses API request.
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
macOS or Linux
Run this command as your normal user:
curl -fsSL https://modelflare.dev/install/codex.sh | sh
When you are already logged in as root on a VPS or container, use:
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:
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 modelflare_api_key.
Step 1: install Codex CLI
Windows PowerShell
irm https://chatgpt.com/codex/install.ps1 | iex
codex --version
Linux
curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex --version
macOS
curl -fsSL https://chatgpt.com/codex/install.sh | sh
codex --version
Step 2: create and enter the .codex directory
Windows PowerShell
New-Item -ItemType Directory -Force "$HOME\.codex" | Out-Null
Set-Location "$HOME\.codex"
Linux
mkdir -p ~/.codex
cd ~/.codex
macOS
mkdir -p ~/.codex
cd ~/.codex
Step 3: create config.toml
Windows PowerShell
$config = @'
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"
[model_providers.modelflare.auth]
command = "powershell.exe"
args = ["-NoProfile", "-NonInteractive", "-Command", "Get-Content -LiteralPath (Join-Path (Join-Path $HOME '.codex') 'modelflare_api_key') -Raw"]
'@
$utf8NoBom = New-Object System.Text.UTF8Encoding -ArgumentList $false
[System.IO.File]::WriteAllText(
(Join-Path (Get-Location) "config.toml"),
$config,
$utf8NoBom
)
Linux
cat > config.toml << 'EOF'
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"
[model_providers.modelflare.auth]
command = "sh"
args = ["-lc", "cat ~/.codex/modelflare_api_key"]
EOF
chmod 600 config.toml
macOS
cat > config.toml << 'EOF'
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"
[model_providers.modelflare.auth]
command = "sh"
args = ["-lc", "cat ~/.codex/modelflare_api_key"]
EOF
chmod 600 config.toml
Step 4: create modelflare_api_key
Before running the command for your system, replace YOUR_MODELFLARE_API_KEY with the complete API key you created in the Modelflare console. The file must contain only the key itself—do not keep the placeholder or add quotes, backticks, or spaces.
Windows PowerShell
$apiKey = @'
YOUR_MODELFLARE_API_KEY
'@
$apiKey = $apiKey.Trim()
$utf8NoBom = New-Object System.Text.UTF8Encoding -ArgumentList $false
[System.IO.File]::WriteAllText(
(Join-Path (Get-Location) "modelflare_api_key"),
$apiKey,
$utf8NoBom
)
$apiKey = $null
Linux
cat > modelflare_api_key << 'EOF'
YOUR_MODELFLARE_API_KEY
EOF
chmod 600 modelflare_api_key
macOS
cat > modelflare_api_key << 'EOF'
YOUR_MODELFLARE_API_KEY
EOF
chmod 600 modelflare_api_key
Step 5: start Codex
Windows 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
cd ~
mkdir -p my-codex-project
cd my-codex-project
codex --model gpt-5.6-sol
macOS
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:
codex --model gpt-5.6-sol
Then ask Codex:
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. |