At a Glance:
CopilotKit is an open-source full-stack SDK for building agent-native applications with generative UI, shared state, and human-in-the-loop workflows, built around the AG-UI protocol for connecting agent logic to user-facing interfaces.
Overview:
CopilotKit is an SDK for building full-stack agentic applications that combine chat interfaces, tool execution, and dynamic UI generation. It connects UI components, backend agents, and tools into a single interaction loop, enabling agents to request user input, render UI directly from backend tool calls, and maintain synchronized state across sessions. The project also develops the AG-UI protocol, an agent-user interaction specification designed to connect agent workflows to user-facing applications through integrations with frameworks like LangGraph and CrewAI. The SDK targets developers building applications where agents and interfaces need to interactively co-create UI and manage multi-step workflows.
Key Decision Points:
Full-stack SDK with React UI: Provides a React-based chat component and hooks, fitting projects that use or are willing to adopt React for the frontend.
Built on the AG-UI protocol: Uses the AG-UI protocol as its foundation, which is also integrated by LangGraph, CrewAI, and other agent frameworks; relevant for teams standardizing on a common interaction protocol.
Agent–UI state sharing: Offers a shared state layer readable and writable by both agents and UI components in real time, supporting workflows that require persistent context across steps.
Human-in-the-loop support: Agents can pause execution to request user input, confirmation, or edits, which is essential for applications requiring approval steps or guided correction.
Claude Code plugin availability: The repository itself can be installed as a Claude Code plugin with multiple skills, providing lifecycle guides for tasks like going to production or scaling to multi-agent systems.
Core Features:
Chat UI: A React-based chat interface with message streaming, tool call support, and agent responses.
Backend Tool Rendering: Allows agents to call backend tools that return UI components rendered directly in the client.
Generative UI: Enables agents to dynamically generate and update UI components at runtime based on user intent and agent state.
Shared State: A synchronized state layer that both agents and UI components can read from and write to in real time.
Human-in-the-Loop: Agents can pause execution to request user input, confirmation, or edits before continuing.
useAgent Hook: A hook that provides direct control over the agent connection, described as a superset of
useCoAgentand built on AG-UI.
Use Cases:
Developers building agent-native applications: For projects where agents need to interactively generate user interfaces, not just return text or structured data.
Applications requiring guided agent workflows: For scenarios where agents must pause for human approval, edits, or additional input during multi-step tasks.
React-based projects integrating agent frameworks: For teams already using or evaluating agent backends like LangGraph or CrewAI that support the AG-UI protocol.
Open-Source Alternative Value:
CopilotKit is distributed under the MIT License and provides an open-source SDK for building full-stack agent applications. The project publishes the AG-UI protocol specification, which defines interaction patterns between agents and user interfaces and has been adopted by multiple agent frameworks including LangChain, AWS, and Microsoft. Developers can embed the SDK’s React components and hooks directly into their applications, use the shared state and human-in-the-loop patterns as defined by the protocol, and reference the repository’s lifecycle guides for tasks such as production deployment and multi-agent scaling.
