CopilotKit/CopilotKit

What it solves

CopilotKit provides a unified SDK to bridge the gap between AI agents and user interfaces. It allows developers to build "agent-native" applications where agents can interact with the app's state, render dynamic UI components, and be deployed across multiple platforms (web, mobile, and chat apps like Slack or Microsoft Teams) without rewriting the agent logic.

How it works

CopilotKit acts as a horizontal layer between the agent backend and the frontend. It uses the AG-UI Protocol to handle the communication between the agent and the user interface. This enables a loop where agents can call backend tools that render UI components in the client, read/write to a shared state layer, and pause execution to request human-in-the-loop confirmation or input.

Who it’s for

Full-stack developers building AI-powered applications who want to integrate complex agentic workflows into React, Angular, Vue, or React Native apps, as well as enterprise chat platforms.

Highlights

• Generative UI: Agents can dynamically create and update UI components at runtime based on user intent.
• Shared State: A synchronized state layer accessible by both the agent and the UI.
• Multi-Channel Support: Deploy the same agent to web, mobile, and chat apps (Slack, MS Teams) via the Channels SDK.
• Human-in-the-Loop: Built-in support for agents to pause and request user input or edits.
• Self-Learning (Early Access): Continuous Learning from Human Feedback (CLHF) allows agents to improve from interactions without model fine-tuning.