| Key features | - Adjustable word generation distance from close to home to makes you reach
- Dig into words by clicking to generate six new words radiating from that one
- Change your question mid-session and reword it with your kept and sparked words
- Taste the stew to see your question rewritten with your choices highlighted
- Ask for ideas that arrive as cards built from your kept and sparked choices
- Turn ideas over to find analogies from different fields
- Get a shortlist of naming candidates or a developed prompt as your finish
- Email your finished brief or shortlist as a one-pager document
- Swipe gestures on phone: left keeps, up sparks, right drops
| - Encrypted vault for AI account credentials on user's computer
- Support for multiple AI providers including ChatGPT, Claude, Gemini, and OpenRouter
- MCP server support to expose site tools for AI access
- Integration with existing AI SDKs (OpenAI, Anthropic, Google Gen AI, Vercel AI SDK, LangChain, LlamaIndex)
- React, Next.js, and plain JavaScript quickstart packages
- Browser support for Chrome, Edge, Brave, and Opera
- Tool result envelope sealing to prevent prompt injection attacks
- Answer guard to control which links appear in AI responses
- 12-word recovery phrase for vault access
- Automatic account switching when plan runs out via Auto mode
|
|---|
| What makes it different | - Recognition is easier than production: react to words instead of articulating ideas from scratch
- Spark steers the generation hardest and is the most powerful control in the app
- Your preferences are demonstrated through actions, not just declared
| - Websites pay nothing for AI inference as it runs on visitors' own accounts
- Visitor credentials stay encrypted in wallet with no key exposure to websites
- Works with existing AI SDKs requiring only one line of code change
- Supports multiple stacks including React, Vue, Svelte, Shopify and WordPress
- Visitor retains control to approve, switch, or revoke site access to their AI account
|