Components as Tools

Let your agent render rich React components directly in the chat by calling them as tools.

Tool-Based Generative UI

Waking up the demo

What is this?

Tool-based Generative UI is the simplest form of Generative UI: you register a React component with useComponent, and CopilotKit exposes it to the agent as a tool. When the agent calls the tool, CopilotKit renders your component inline in the chat, passing the tool's arguments straight through as typed props.

Unlike tool rendering, which wraps a real backend tool in a custom UI, tool-based GenUI is the component. There is no handler, no user interaction, no server-side execution. The agent decides when to show it, populates the data, and CopilotKit renders it.

When should I use this?

Use useComponent when you want to:

  • Display rich UI (cards, charts, tables, dashboards) inline in the chat
  • Show structured data the agent has derived from its reasoning
  • Render previews, status indicators, or visual summaries
  • Let the agent present information beyond plain text

For components that need user interaction, see Human-in-the-loop. For operational transparency around a real backend tool, see Tool rendering.

How it works in code

Forward browser tools to Claude

Frontend tools registered with useFrontendTool arrive in the AG-UI run input. This demo's server calls the Anthropic Messages API directly for runs that carry frontend tools, so it converts each AG-UI tool definition into a Messages API tool schema first.

The published ClaudeAgentAdapter from @ag-ui/claude-agent-sdk needs no such conversion. It registers each frontend tool on its own in-process MCP tool server as mcp__ag_ui__<name>, allows it, and ends the run when Claude calls it so the browser can execute the tool. Pass the run input to the adapter and frontend tools work with no extra wiring.

agent_server.ts
function buildTools(tools: RunAgentInput["tools"]): Anthropic.Tool[] {
  if (!tools || tools.length === 0) return [];

  return tools.map((tool) => {
    let inputSchema: Anthropic.Tool.InputSchema = {
      type: "object",
      properties: {},
    };
    if (tool.parameters) {
      try {
        const parsed =
          typeof tool.parameters === "string"
            ? JSON.parse(tool.parameters)
            : tool.parameters;
        inputSchema = parsed as Anthropic.Tool.InputSchema;
      } catch (parseErr) {
        // Don't silently swap in an empty schema — Claude will then accept
        // any input shape, which compounds whatever caller bug produced
        // the malformed JSON. Warn loudly so the tool definition gets
        // fixed instead of being papered over.
        const message =
          parseErr instanceof Error ? parseErr.message : String(parseErr);
        console.warn(
          `[agent_server] failed to parse tool.parameters for ${tool.name}; using empty schema. error=${message}`,
        );
      }
    }
    return {
      name: tool.name,
      description: tool.description ?? "",
      input_schema: inputSchema,
    };
  });
}

The built-in agent needs no backend step. Import the hook and Zod in the component that registers the tool:

import { useComponent } from "@copilotkit/react-core/v2";
import { z } from "zod";

useComponent takes a name, a Zod schema for its props, and the component to render. The hook registers it as a frontend tool, so the agent can call it. The schema becomes the tool's parameters, which tell the model which arguments to send.

parameters is optional, but leaving it out advertises the tool with an empty parameter schema ({ "type": "object", "properties": {} }). The model then has nothing to fill in, so it calls the tool with no arguments and your component renders with no props. Pass a schema for any component that needs data.

Define the props of your component as a Zod schema:

bar-chart.tsx
import { z } from "zod";

export const barChartPropsSchema = z.object({
  title: z.string().describe("Chart title"),
  description: z.string().describe("Brief description or subtitle"),
  data: z.array(
    z.object({
      label: z.string(),
      value: z.number(),
    }),
  ),
});

Pass the schema and your component to useComponent:

page.tsx
  useComponent({
    name: "render_bar_chart",
    description: "Display a bar chart with labeled numeric values.",
    parameters: barChartPropsSchema,
    render: BarChart,
  });

BarChart is your own React component. It reads only its props and can stream in as the agent fills the payload. The demo builds it with Recharts. It does not know anything about CopilotKit.

The name you pass to useComponent is what the agent sees as the tool name. Make it a verb like render_bar_chart or show_weather so the LLM reliably picks it when the user asks for that visualization.

To check it, ask the agent for a bar chart of quarterly sales. The bar chart renders inline in the chat with the data from the agent.

Rendering in a headless chat

CopilotKit's built-in chat components render registered components for you. A headless or custom chat renders the message list itself, so no tool call renders unless you render it. The component is registered and the agent calls it, but the chat stays empty.

Render the tool calls on each assistant message with CopilotChatToolCallsView:

import { CopilotChatToolCallsView } from "@copilotkit/react-core/v2";

<CopilotChatToolCallsView message={assistantMessage} messages={allMessages} />;

It looks up the sibling tool-role message for each tool call and hands both to the registered renderer. For finer placement, call useRenderToolCall() and render each tool call yourself. See Headless UI.

How is this guide?