End-to-End Type Safety
Every exposed API that allows custom types propagates them end-to-end, including to the frontend. No more type mismatches!
Plug-and-Play
First-class support for AI tool calls with real-time updates, interactive callbacks, and structured response types.
Automated State Management
Client-side state is managed through signals and avoids redundant re-instancing of identical data for optimal performance.
Framework Agnostic
Adapters can be written for any LLM backend or frontend framework. Currently supports Langchain and React with more coming soon.
Client-Side Tool Previews
Split client/server argument schemas allow the client to preview partial arguments before they are sent to the tool function.
Minimized Re-rendering
Built on tRPC primitives with a signals-based state approach, optimizing data flow and ensuring minimal re-renders.
Scroll further...