AI models: use your own Claude, ChatGPT or Gemini subscription
Island Pilot runs on your own AI subscription: Claude, ChatGPT, Gemini and more. We never resell a model and take no cut of your calls. You pick each agent’s model and reasoning effort, and you can switch the whole team to another engine in one go.
Claude, ChatGPT, Gemini: supported AI providers
| Provider | Gateway | Sign-in |
|---|---|---|
| Claude (Anthropic) | Claude Code | Your Claude account |
| ChatGPT (OpenAI) | Codex | Your ChatGPT account |
| Gemini (Google) | Gemini CLI | Your Google account |
| OpenCode Zen and Go | OpenCode | Your OpenCode keys |
| Grok (xAI) | xAI API | Your xAI API key |
Island Pilot works through the providers’ official command-line tools. The store detects the one you need, installs it if necessary, and the AI accounts tab shows the sign-in command (Gemini CLI needs Node.js 22 or later). Keys (OpenCode, xAI) stay in the Windows credential vault; other providers’ sessions stay managed by their own tool.
A model and reasoning effort for every agent
Each agent profile keeps its own provider, model and reasoning effort. So you can give the Verse Coder’s Verse work to a strong model and a tidy-up task to a faster one.
- Effort is set with a slider, in the profile and in the conversation. Only the levels the exact model offers are shown; an unknown model stays on Auto.
- When the orchestrator delegates, the specialist keeps its model and effort if they are set explicitly.
- A specialist left on Automatic inherits the orchestrator’s model and reasoning.
See the AI agent team for what each specialist does.
Change engine in one go
Ran out of quota, or want to try another subscription? In Settings › Agents, “Change engine” applies a provider, model and effort to all the agents you tick at once. Optionally, the change extends to their conversations and to conversations without a profile.
You don’t lose the thread: the project’s context doesn’t live inside a model, but in your files, the agents’ reports and the project To-Do. The new engine picks up from there.
The tools stay Island Pilot’s
Whatever the model, actions go through Island Pilot’s checked tools: file reads and writes with hash checks, UEFN, Blender, Meshy and your own MCP servers. The command-line tools’ native tools are turned off.
One exception: when the profile allows web search, the search built into Claude Code, Codex or Gemini is turned on, since it runs on the provider’s servers. Every query is logged in Activity.
What you pay, and to whom
- Island Pilot
- One payment, no subscription, see pricing
- AI calls
- On your subscription or credits, at your provider’s price
- Image generation
- With your ChatGPT account, or a Gemini or xAI API key
- Opening a project or conversation
- No paid call
FAQ
Can I use my Claude or ChatGPT subscription with UEFN?
Yes. Island Pilot connects to your account through Claude Code or Codex, and your agents work in your UEFN project with that model. You pay your subscription directly to its provider.
Does Island Pilot resell model access?
No, never. Island Pilot is software you pay for once; requests only go to the services you connect, with your own accounts.
Do I need to install anything besides Island Pilot?
Your provider’s command-line tool (Claude Code, Codex, Gemini CLI or OpenCode). The store detects it and installs it for you if needed; Grok only needs your xAI API key.
Can I mix providers in the same team?
Yes. Each agent has its own provider and model. The orchestrator can run on Claude while a specialist uses ChatGPT, for example.
What happens when my quota runs out?
The provider refuses new calls until it resets. You can move your agents to another subscription with “Change engine”, without losing the project’s context.