Documentation
One key, two compatible wire protocols, 15 frontier models. Read the quickstart or explore the API reference.
Coding agents · Codex
Codex (by ChatGPT)
Codex is OpenAI's agentic coding CLI. Point it at the Selora gateway from a single config.toml provider block — set a base URL and your key, then launch.
Create an API key on the API Keys page and export it as $SELORA_API_KEY before continuing.
Install Codex
Install the Codex CLI, then verify it is on your PATH.
npm install -g @openai/codexYou can also install via homebrew (brew install codex) or the official installer. Codex needs an OpenAI-compatible gateway.
Add the gateway as a provider
Open ~/.codex/config.toml and add a provider block that points at the gateway. Point the base URL at the OpenAI-compatible endpoint and set the env var that carries your Selora key.
[model_providers.selora]
name = "Selora"
base_url = "https://api.selora.lol/v1"
env_key = "SELORA_API_KEY"
wire_api = "responses"
requires_openai_auth = false
[profiles.selora]
model = "gpt-6-astra"
model_provider = "selora"Set SELORA_API_KEY in your shell. wire_api = "responses" is what current Codex builds require for custom providers — the official config reference lists responses as the only supported value, and it targets /v1/responses. "chat" (/v1/chat/completions) still works but only on older pinned builds, so treat it as legacy. Both wires carry full tool-calling (tools, tool_choice, role:"tool" results, streaming tool_calls deltas), so Codex works as a real agent, not just plain chat. If you use CCSwitch to manage this config, see the CCSwitch guide — it has known bugs that break the generated config, and the selora-setup.mjs script fixes them in one command. Reserved provider names like openai, ollama and lmstudio take precedence and cannot be overridden.
Run it
Launch Codex against the Selora profile. It will call models over the gateway.
export SELORA_API_KEY="sk-gw-..."
cd /path/to/project
codex --profile selora