Codex CLI / App

Use this for OpenAI groups such as Codex Plus Pool and Codex Pro Pool.

This configuration applies to Codex CLI, Codex App, and the VS Code extension.

config.toml

Put the following content at the beginning of config.toml.

macOS / Linux

# ~/.codex/config.toml
model_provider = "codex-everywhere"
model = "gpt-6-astra"
model_reasoning_effort = "high"
disable_response_storage = true
supports_websockets = false

[model_providers.codex-everywhere]
name = "codex-everywhere"
base_url = "https://codex-everywhere.com"
wire_api = "responses"
requires_openai_auth = true

[features]
goals = true

Windows

# %userprofile%\.codex\config.toml
model_provider = "codex-everywhere"
model = "gpt-6-astra"
model_reasoning_effort = "high"
disable_response_storage = true
supports_websockets = false

[model_providers.codex-everywhere]
name = "codex-everywhere"
base_url = "https://codex-everywhere.com"
wire_api = "responses"
requires_openai_auth = true

[features]
goals = true

auth.json

macOS / Linux

{
  "OPENAI_API_KEY": "<your-ce-api-key>"
}

Windows

{
  "OPENAI_API_KEY": "<your-ce-api-key>"
}

1M context window

To enable a 1M context window, update these top-level entries in config.toml, before any section headers:

model = "gpt-6-astra"
model_context_window = 1000000
model_auto_compact_token_limit = 900000