Codex

Add opengateway as a Codex CLI provider profile in ~/.codex/config.toml — run codex --profile opengateway, your default provider stays untouched.

Copy this to your agent

Paste it into any AI agent. It sets up the codex --profile opengateway profile and saves it as the opengateway-codex skill.

Run it once, nothing saved#

Not ready for either? Copy the command instead — the whole provider goes in as -c overrides, nothing is written anywhere, and the next plain codex run is exactly as before:

OPENGATEWAY_API_KEY="YOUR_OPENGATEWAY_API_KEY" codex \
  -c 'model_provider="opengateway"' \
  -c 'model="z-ai/glm-5.2-ultrafast"' \
  -c 'model_providers.opengateway.name="opengateway"' \
  -c 'model_providers.opengateway.base_url="https://apis.opengateway.ai/v1"' \
  -c 'model_providers.opengateway.env_key="OPENGATEWAY_API_KEY"' \
  -c 'model_providers.opengateway.wire_api="responses"'

Set it up by hand#

1. Get an API key#

opengateway.ai → API Keys → Create Key. Export it in your shell profile (~/.zshrc or ~/.bashrc):

export OPENGATEWAY_API_KEY="YOUR_OPENGATEWAY_API_KEY"

2. Add the profile#

Add to ~/.codex/config.toml — any model ID from the catalog works; z-ai/glm-5.2-ultrafast is just the example:

# opengateway-codex:start
[model_providers.opengateway]
name = "opengateway"
base_url = "https://apis.opengateway.ai/v1"
env_key = "OPENGATEWAY_API_KEY"
wire_api = "responses"
# opengateway-codex:end

Then put the profile in its own file, ~/.codex/opengateway.config.toml:

model_provider = "opengateway"
model = "z-ai/glm-5.2-ultrafast"

The marker comments make the config.toml block ours — undoing the setup is deleting everything between them plus the profile file, nothing else changes. On an old Codex that errors on the separate profile file, use a [profiles.opengateway] table inside the markers instead — recent Codex rejects that legacy table, which is why the file is the default.

wire_api = "responses" matters — Codex talks the Responses API (/v1/responses), which opengateway serves natively. To make opengateway the default, set top-level model and model_provider too.

3. Verify#

codex --profile opengateway exec "Reply with OK"

A response means it works — the call is already on the dashboard with tokens, cost, and latency.