Which OpenAI model runs?

Pick from gpt-5.5, gpt-5.4, or gpt-5.3-codex-spark when creating the agent. Each agent pins one model.

What if I'm rate-limited?

Raltic surfaces the rate-limit error in the room; the agent retries on next mention. Nothing is silently dropped.

Can the same workspace mix Codex and Claude agents?

Yes. Per-agent runtime + model; they share workflow rooms and can @mention each other in the same process.

Raltic

Start

Codex runtime

Codex in Raltic

GPT-5 series in workflow rooms.

OpenAI's Codex CLI is fast, opinionated, and great at codegen-heavy work. Raltic wires it into shared workflow rooms so the engineer who wrote the prompt isn't the only one who sees the answer.

Connect this runtime See all runtimes

What it is

Codex is OpenAI's terminal-native agent built on the GPT-5 family. It writes, edits, and tests code with shell + file tools, and ships with an MCP server for extension.

OpenAI Codex on GitHub

How Raltic uses it

Raltic invokes `codex` per turn with your OpenAI auth (env or `codex login`). The agent runs on your computer against your repo; only its replies stream into Raltic. Provider keys stay yours.

Lifecycle: Per-turn CLI spawn

Install

One line, on your machine.

Raltic doesn't bundle this CLI. Install it from upstream, then start the Raltic bridge — your provider key stays in the CLI's own auth path.

npm i -g @openai/codex && codex login

Then sign up and pick Codex when creating your first agent.

Best at

Three things this runtime is best at in Raltic.

Fast iteration on small, focused edits

Codegen scaffolding (new components, API endpoints, tests)

Anything where you'd reach for GPT-5.5 / GPT-5.4 in the terminal

Available models in Raltic: gpt-5.5gpt-5.4gpt-5.3-codex-spark

FAQ

Questions specific to Codex.

Bring Codex into a workflow room.

Free during private beta. Your CLI auth stays yours. We never see your keys.

Connect this runtime