OpenAI Codex

OpenAI Codex is OpenAI's agentic coding tool. You can configure it to use the Skytells API as its model provider, giving you access to any model in the Skytells catalog while keeping request traffic, usage metrics, and spend visible in the Skytells Console.

By pointing Codex at Skytells, you can:

• Route requests through any model supported by Skytells — OpenAI, Anthropic, Google, and more
• Monitor traffic and token consumption in the Skytells Console
• Switch between models without touching your application code
• Use a single API key across all your agentic tools

Configure OpenAI Codex

1. Install OpenAI Codex CLI

Follow the installation instructions on the OpenAI Codex repository to install the Codex CLI tool.

2. Configure your environment variable

Set your Skytells API key in your shell configuration file:

export SKYTELLS_API_KEY="your-skytells-api-key"

After adding the variable, reload your shell configuration:

source ~/.zshrc  # or source ~/.bashrc

3. Set up the Codex config file

Open ~/.codex/config.toml and add the following configuration:

[model_providers.skytells]
name    = "Skytells AI"
base_url = "https://api.skytells.ai/v1"
env_key  = "SKYTELLS_API_KEY"
wire_api = "responses"

[profiles.skytells]
model_provider = "skytells"
model          = "openai/gpt-5.3-codex"

This configuration:

• Registers a model provider named skytells pointing to the Skytells API
• Reads your SKYTELLS_API_KEY environment variable for authentication
• Creates a skytells profile that uses this provider
• Sets openai/gpt-5.3-codex as the default model for this profile
• Uses wire_api = "responses" to match the OpenAI Responses API wire format, which Skytells fully supports

4. Run Codex

Start Codex with the skytells profile:

codex --profile skytells

All requests are routed through the Skytells API. You can monitor usage and request traces from the Skytells Console.

5. (Optional) Use a different model

To switch models, update the model field in your profile:

[profiles.skytells]
model_provider = "skytells"
model          = "anthropic/claude-sonnet-4.6"

Any model available through the Skytells API can be used here. Refer to the Models reference for the full list.

6. (Optional) Define multiple profiles

You can define several profiles in the same config file and switch between them at runtime:

[model_providers.skytells]
name     = "Skytells AI"
base_url = "https://api.skytells.ai/v1"
env_key  = "SKYTELLS_API_KEY"
wire_api = "responses"

[profiles.skytells]
model_provider = "skytells"
model          = "openai/gpt-5.3-codex"

[profiles.fast]
model_provider = "skytells"
model          = "openai/gpt-4o-mini"

[profiles.reasoning]
model_provider = "skytells"
model          = "openai/o3"

[profiles.claude]
model_provider = "skytells"
model          = "anthropic/claude-sonnet-4.6"

# Or try other models:
# model = "google/gemini-3-flash"
# model = "openai/o3"

Switch between profiles using the --profile flag:

codex --profile skytells    # default Codex model
codex --profile fast        # lightweight, low-latency
codex --profile reasoning   # extended reasoning
codex --profile claude      # Anthropic Claude

Notes