CODING CLIENTS

Codex CLI

Use your Yugos project key with the Responses API.

1. Create a key

At launch, create a project key in API keys (coming soon), then set YUGO_API_KEY in your terminal environment. Keep it out of source control.

2. Add a separate configuration

Create ~/.codex-yugo/config.toml with the configuration below. This keeps your existing Codex settings available.

model = "yugo"
model_provider = "yugo"
web_search = "disabled"
model_reasoning_effort = "low"
model_reasoning_summary = "none"

[features]
enable_request_compression = false

[model_providers.yugo]
name = "Yugos API"
base_url = "https://bitmind--yugo-api-gateway-fastapi-app.modal.run/v1"
env_key = "YUGO_API_KEY"
wire_api = "responses"
requires_openai_auth = false
supports_websockets = false
request_max_retries = 0
stream_max_retries = 0

These connection fields follow the official Codex configuration reference. The settings above are the tested Yugos staging configuration.

3. Start Codex

export YUGO_API_KEY="<your project key>"
CODEX_HOME="$HOME/.codex-yugo" codex --model yugo

Use your usual local permissions and sandbox settings. Tools execute on your machine. Start a fresh conversation when changing model configurations; returned reasoning state stays with its original model.

Inspect usage

Open Usage & spend (coming soon) to see Gemini costs. If a stream fails, inspect its request receipt before retrying. Automatic retries are disabled in this setup to make that decision explicit. See compatibility for current limits.