Connect an AI provider
Use your Claude or ChatGPT subscription through the official CLIs, or an API key. Here is how they differ.
EaseKit designs animations by chatting with an AI. You can connect four kinds of provider and switch between them any time in Settings → Providers. Subscriptions are the easiest start; API keys unlock every feature and bill per use.
| Provider | How it signs in | Billing |
|---|---|---|
| Claude subscription (Claude Code) | Sign in once in the browser through Claude Code | Your Claude Pro or Max plan |
| ChatGPT subscription (Codex) | Sign in once in the browser through Codex | Your ChatGPT plan |
| Claude (API key) | Paste an Anthropic API key | Pay per use through the API |
| OpenAI / Codex (API key) | Paste an OpenAI API key | Pay per use through the API |
Use a subscription
Signing in with a subscription uses your plan's credits through the official command-line tools, so you don't need an API key.
- Open Settings → Providers (or the Connect AI step in setup).
- Pick Claude subscription or ChatGPT subscription.
- If the CLI isn't installed yet, click Install in Terminal. EaseKit types the install command for you (or copies it, so you can paste it with ⌘V and press Return).
- Click Sign in. A Terminal window opens; confirm once in your browser.
- When the CLI shows as found and signed in, choose a Model and click Use to make it the provider in use.
Chats resume their Claude Code or Codex session, so each message only sends what's new and your plan lasts longer. If you have both CLIs, turn on Continue with the other plan at the limit: when one plan runs out, the same message continues on the other. Short hiccups are retried once.
Use an API key
- Open Settings → Providers.
- Paste your key into Claude (API key) or OpenAI / Codex (API key) and click Save.
- Choose a Model. Click the reload button to load the latest models from the provider.
- Set Default provider to the one you want new chats to use.
Keys are stored in the macOS Keychain. EaseKit also reads ANTHROPIC_API_KEY and OPENAI_API_KEY from the environment as a fallback.
AI power and responses
These settings live in Settings → Providers and apply to every provider.
| Setting | Options | What it does |
|---|---|---|
| AI power | Eco, Balanced (default), Max | Eco: brief reasoning, an outline instead of the full animation, targeted patches. Balanced: the full animation up to medium size, skills as needed. Max: deep reasoning, everything in context, more review rounds |
| Reasoning effort | Match AI power, Low, Medium, High, Very high | How long the model thinks before it answers |
| Show reasoning summary | On / off | Shows what the model is thinking while it works |
| Claude: refusal fallback (server-side) | On / off | For Claude models that support it, another model takes over when a request is declined |
| Max response length | 16k, 32k, 64k tokens | Raise it if long answers get cut off |
| History in context | Compact, Balanced, Large | How much of the conversation is replayed each turn (about 60k, 240k or 900k characters) |
Figma
The same page has a Figma section for a personal access token (scope File content: Read-only, kept in your Keychain), used by Import from Figma.
Works with
- Prompts that work: what to ask once you're connected.
- How the AI edits your animation: what the model sees each turn.
- Privacy: what leaves your Mac.
Troubleshooting
| If this happens | Try this |
|---|---|
| Install in Terminal only copies the command | Terminal control isn't allowed. Paste the command in Terminal with ⌘V and press Return. |
| Sign-in didn't finish | Click Sign in again and confirm in the browser window that opens. |
| The model list is empty or old | Click the reload button next to Model. |
| “Usage limit reached” | Wait for your plan to reset, switch to an API key, or turn on Continue with the other plan at the limit. |