useHumanInTheLoop

The useHumanInTheLoop hook enables human approval and interaction workflows.


This is the v1 useHumanInTheLoop, which takes an array of parameter definitions. It is still supported, but we recommend migrating to the v2 useHumanInTheLoop (from @copilotkit/react-core/v2), which uses Zod schemas for parameters.

useHumanInTheLoop pauses AI execution to request human input or approval. When the AI calls this tool, it stops and waits for the user to respond through your custom UI before continuing. This is essential for sensitive operations, confirmations, or collecting information that only the user can provide.

Unlike useFrontendTool, there's no handler function. Instead, your render function receives a respond callback that sends the user's input back to the AI. The AI execution remains paused until respond is called, making this a true blocking interaction.

Usage

Simple Confirmation Example

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

useHumanInTheLoop({
  name: "confirmDeletion",
  description: "Ask user to confirm before deleting items",
  parameters: [
    {
      name: "itemName",
      type: "string",
      description: "Name of the item to delete",
      required: true,
    },
    {
      name: "itemCount",
      type: "number",
      description: "Number of items to delete",
      required: true,
    },
  ],
  render: ({ args, status, respond, result }) => {
    if (status === "executing" && respond) {
      return (
        <div className="p-4 border rounded">
          <p>Are you sure you want to delete {args.itemCount} {args.itemName}(s)?</p>
          <div className="flex gap-2 mt-4">
            <button
              onClick={() => respond({ confirmed: true })}
              className="bg-red-500 text-white px-4 py-2 rounded"
            >
              Delete
            </button>
            <button
              onClick={() => respond({ confirmed: false })}
              className="bg-gray-300 px-4 py-2 rounded"
            >
              Cancel
            </button>
          </div>
        </div>
      );
    }

    if (status === "complete" && result) {
      return (
        <div className="p-2 text-sm text-gray-600">
          {result.confirmed ? "Items deleted" : "Deletion cancelled"}
        </div>
      );
    }

    return null;
  },
});

Complex Input Collection Example

import { useHumanInTheLoop } from "@copilotkit/react-core";
import { useState } from "react";

useHumanInTheLoop({
  name: "collectUserPreferences",
  description: "Collect detailed preferences from the user",
  parameters: [
    {
      name: "context",
      type: "string",
      description: "Context for why preferences are needed",
      required: true,
    },
    {
      name: "requiredFields",
      type: "string[]",
      description: "Fields to collect",
      required: true,
    },
  ],
  render: ({ args, status, respond }) => {
    const [preferences, setPreferences] = useState({
      theme: "light",
      notifications: true,
      language: "en",
    });

    if (status === "executing" && respond) {
      return (
        <div className="p-4 border rounded">
          <h3 className="font-bold mb-2">{args.context}</h3>
          <form onSubmit={(e) => {
            e.preventDefault();
            respond(preferences);
          }}>
            <button
              type="submit"
              className="bg-blue-500 text-white px-4 py-2 rounded"
            >
              Save Preferences
            </button>
          </form>
        </div>
      );
    }

    return null;
  },
});

Best Practices

  1. Always check for the respond function before rendering interactive elements
  2. Handle all status states to provide good user feedback
  3. Validate user input before calling respond
  4. Provide clear instructions in your UI about what input is expected
  5. Consider timeout scenarios for time-sensitive operations

Migration from useCopilotAction

If you're migrating from useCopilotAction with renderAndWaitForResponse:

// Before with useCopilotAction
useCopilotAction({
  name: "confirmAction",
  parameters: [
    { name: "message", type: "string", required: true },
  ],
  renderAndWaitForResponse: ({ args, respond, status }) => {
    return (
      <ConfirmDialog
        message={args.message}
        onConfirm={() => respond(true)}
        onCancel={() => respond(false)}
        isActive={status === "executing"}
      />
    );
  },
});

// After with useHumanInTheLoop
useHumanInTheLoop({
  name: "confirmAction",
  parameters: [
    {
      name: "message",
      type: "string",
      description: "The message to display",
      required: true,
    },
  ],
  render: ({ args, respond, status }) => {
    if (status === "executing" && respond) {
      return (
        <ConfirmDialog
          message={args.message}
          onConfirm={() => respond(true)}
          onCancel={() => respond(false)}
          isActive={true}
        />
      );
    }
    return null;
  },
});

The main differences are:

  1. The property is called render instead of renderAndWaitForResponse
  2. You need to check for the respond function's existence

Parameters

Prop

Type

Prop

Type

Prop

Type

Prop

Type

Prop

Type