Which AI models and coding agents does Draftbit support?

Two coding agents run inside Draftbit today, Claude Code from Anthropic and Codex from OpenAI. Google Gemini is on the roadmap and is not available yet.

Each thread pairs an agent with a model. Anthropic’s Claude models run through Claude Code, and OpenAI’s models run through Codex. Models from other vendors reach Draftbit through OpenRouter. That catalog currently includes models from Google, xAI, DeepSeek, Meta, Moonshot AI, MiniMax, Qwen, and Z.ai, and it changes as vendors ship and retire models. OpenRouter models run through Codex, and some can run through Claude Code as well. The model picker in the chat shows what your account can use right now, and Agent Customization lets you set a default agent and model.

Your plan matters when Draftbit credits pay for the thread. On the Free plan, credit-funded threads use Codex with supported OpenAI models. Paid plans get the full agent and model choice with credits. If you connect your own OpenAI, Anthropic, or OpenRouter key or a ChatGPT or Claude subscription, which every plan allows, model availability follows what your provider account permits rather than a Draftbit list.

Adding AI to the app you are building is a separate question. Your app can call any model API you choose, and the agent can help you write that code. Keep the provider key in a sensitive environment variable. Plain text EXPO_PUBLIC variables are bundled into the app, so a private key must never go there.