AI Integrator runtime
Run 40 models in Codex, Claude Code, or OpenCode on your AI Integrator plan.
- Choose it as the runtime
- Pick AI Integrator in the runtime menu, then one of 40 models. You don't need an OpenAI or Anthropic account.
- Use the program that fits the model
- OpenAI models run in Codex, Anthropic models in Claude Code, and the rest in OpenCode. If a model's own program isn't installed, OpenCode runs it.
- Pay from your plan
- Requests count toward the same monthly allowance as Chat. Free includes the standard models, and Pro and Max include them all.
- Keep your privacy settings
- Zero data retention and no training are on by default, and the account applies them to every request. A model that can't keep zero data retention stays locked until you turn that setting off.
What runs where
- OpenAI models, such as GPT-6 Astra, run in Codex.
- Anthropic models, such as Claude Opus 5.5, run in Claude Code.
- Gemini, Grok, DeepSeek, Qwen, GLM, MiniMax, Mistral, and other models run in OpenCode.
- If a model's own program isn't installed, OpenCode runs it.
- Usage counts toward the same allowance as Chat.
- A model that can't keep zero data retention stays locked while that setting is on.
Questions
No. The runtime uses your AI Integrator plan usage. Codex, Claude Code, or OpenCode must be installed on your computer, because one of them runs the model.
40 models from the Chat model list, including OpenAI, Anthropic, Google, Grok, Meta, DeepSeek, Alibaba, and Mistral models. The model menu shows which ones your plan and privacy settings allow.
Codex runs OpenAI models, Claude Code runs Anthropic models, and OpenCode runs the rest. If a model's own program isn't installed, OpenCode runs it. If none of them is installed, the app asks you to install OpenCode.
No. Requests pass through your account to the model host, and no prompt or reply is stored. We keep the model name, token counts, and cost for billing.
No. Your repository stays on your computer. Each request carries the prompt and the files and command output the agent chose to include, through your account to the model host.
Yes. Both are on by default. The account applies your settings to every request, and a program or a prompt can't turn them off. A model that can't keep zero data retention stays locked until you turn that setting off.