Codex CLI

Codex CLI uses OpenAI-style APIs. Configure a custom provider that points to ModRouter.

Endpoint

Codex should use an OpenAI-compatible base URL that includes /v1:

https://api.modrouter.online/openai/v1

Set MODROUTER_BASE_URL in the shell or config-management system that writes your Codex config.

Setup

Put provider configuration in your user-level Codex config, commonly ~/.codex/config.toml:

export OPENAI_API_KEY="${MODROUTER_API_KEY}"

model = "openai/gpt-4o-mini"
model_provider = "modrouter"

[model_providers.modrouter]
name = "ModRouter"
base_url = "https://api.modrouter.online/openai/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"

If you route Codex to non-OpenAI models, disable WebSocket-only behavior in Codex versions that support that switch:

[model_providers.modrouter]
name = "ModRouter"
base_url = "https://api.modrouter.online/openai/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"
supports_websockets = false

Models

Use explicit ModRouter model IDs:

codex --model openai/gpt-4o-mini
codex --model anthropic/claude-sonnet-4-5
codex --model gemini/gemini-2.5-pro

Codex relies on tool calling for editing and shell workflows. Pick models that can reliably produce tool calls in OpenAI-compatible format.

CLI Agents & Editors

MCP Gateway

Endpoint

Setup

Models

​Endpoint

​Setup

​Models

Endpoint

Setup

Models