Programmatic Control

Drive agent runs from code in a React Native app, with no chat UI required.

What is this?

Programmatic control is what you reach for when you want to drive an agent run from code rather than from a chat composer: a button, a form, a push notification, a voice transcript, or a background task. CopilotKit gives you three primitives that cover every trigger:

  • agent.addMessage(...): add a message to the conversation without running the agent. Pair it with copilotkit.runAgent({ agent }) when the message should start a turn.
  • copilotkit.runAgent({ agent }): the same entry point the prebuilt chat calls. It runs frontend tools, follow-up runs, and the subscriber lifecycle.
  • agent.subscribe(subscriber): a low-level AG-UI event subscription (onCustomEvent, onRunStartedEvent, onRunFinalized, onRunFailed, and more). Use it with copilotkit.runAgent({ agent, resume }) for standard AG-UI interrupts, or with copilotkit.runAgent({ agent, forwardedProps: { command: { resume, interruptEvent } } }) for the on_interrupt custom events that the LangGraph adapter sends.

When should I use this?

Use programmatic control when you want to:

  • Start agent runs from buttons, forms, or other views
  • Run one of your frontend tools directly, without an LLM turn
  • Build agent features without a chat screen (to build your own chat, see Headless UI)
  • Read agent state and results from code
  • Send a speech-to-text transcript as a user turn

Sending a message from code

Install the OpenAI Agents AG-UI adapter

pip install ag-ui-openai-agents openai-agents

Wire the agent once, and chat and code share it

copilotkit.runAgent, agent.subscribe and agent.addMessage drive runs against the same AG-UI endpoint your chat UI posts to, so there is no second backend to set up. Mount the OpenAIAgentsAgent the way any chat demo does, and every entry point sees the same model, tools and instructions.

agent_server.py
app = FastAPI(title="OpenAI Agents SDK Agent Server", lifespan=lifespan)
for name, agent in AGENTS.items():
    add_openai_agents_fastapi_endpoint(app, agent, f"/{name}", include_health=False)

The server keeps no sessions. Every run carries the thread's messages from the client, so a headless run on the same agent continues its thread, and a fresh threadId starts a new one.

Headless interrupts come from approval-gated tools

The headless interrupt-resolver pattern (read the interrupt off the agent, answer it with a resume) works with a server tool marked needs_approval=True. The run ends with an interrupt for the call, and a resume of {"approved": true} runs it (see the human-in-the-loop setup). To ask the user from code without a backend tool, use the frontend-tool Promise pattern: register the tool with useFrontendTool and an async handler, and resolve its promise when your own UI has the answer.

Add a user message with agent.addMessage, then call copilotkit.runAgent({ agent }). Call copilotkit.stopAgent({ agent }) to cancel a run in flight.

src/AgentTrigger.tsx
import { Button, Text, View } from "react-native";
import { useAgent, useCopilotKit } from "@copilotkit/react-native/headless";

export function AgentTrigger({ agentId }: { agentId: string }) {
  const { agent } = useAgent({ agentId });
  const { copilotkit } = useCopilotKit();

  const run = async () => {
    if (agent.isRunning) return;

    agent.addMessage({
      id: `user-${Date.now()}`,
      role: "user",
      content: "Summarize the latest sales data",
    });

    try {
      await copilotkit.runAgent({ agent });
    } catch (error) {
      console.error("CopilotKit runAgent failed:", error);
    }
  };

  const last = agent.messages[agent.messages.length - 1];

  return (
    <View>
      <Button title="Run agent" onPress={run} disabled={agent.isRunning} />
      <Button
        title="Stop"
        onPress={() => copilotkit.stopAgent({ agent })}
        disabled={!agent.isRunning}
      />
      {last?.role === "assistant" && last.content ? <Text>{last.content}</Text> : null}
    </View>
  );
}

Tap Run agent. The agent's reply streams into agent.messages, and the last assistant message shows under the buttons.

The message id comes from a timestamp. The React Native polyfills install crypto.getRandomValues, not crypto.randomUUID.

copilotkit.runAgent() vs agent.runAgent()

Both methods start the agent, but they work at different levels:

  • copilotkit.runAgent({ agent }): the default. It runs the full lifecycle: it executes frontend tools, handles follow-up runs, and sends errors to the subscriber system and the provider's onError.
  • agent.runAgent(options): the low-level method on the agent instance. It sends the request to the runtime but does not execute frontend tools or chain follow-ups. Use it only when you need direct control.

Running a tool without the model

copilotkit.runTool() runs a registered frontend tool directly. The handler runs, the tool's render shows in the chat, and both the tool call and its result are added to agent.messages. No LLM turn happens unless you ask for one.

const { copilotkit } = useCopilotKit();

const { result, error } = await copilotkit.runTool({
  name: "filter_tasks",
  parameters: { status: "open" },
  // false (default): no follow-up. "generate": the agent responds to the result.
  followUp: false,
});

The tool must be registered with useFrontendTool somewhere under the provider. See Frontend Tools.

Subscribing to agent events

agent.subscribe(subscriber) returns { unsubscribe }. The subscriber accepts every AG-UI lifecycle callback: onCustomEvent, onRunStartedEvent, onRunFinalized, onRunFailed, and the streaming deltas. Use it to drive your own progress UI, to send events to analytics, or to catch pause and resume events and answer them, as shown below.

Subscribe in an effect and unsubscribe in its cleanup:

import { useEffect } from "react";

useEffect(() => {
  const { unsubscribe } = agent.subscribe({
    onRunStartedEvent: () => setStatus("working"),
    onRunFinalized: () => setStatus("done"),
    onRunFailed: () => setStatus("failed"),
  });
  return unsubscribe;
}, [agent]);

useAgent can swap in a new agent object once the runtime connects, so keep agent in the effect's dependencies.

Resolving a pause from a button

Interrupt-style pause and resume isn't available on this framework. The headless interrupt pattern needs a runtime with a native interrupt(...) primitive (LangGraph) or a Promise-resolving frontend-tool path. For all other integrations, pause with useHumanInTheLoop instead. It works on every framework that supports tool calls. The agent.addMessage, copilotkit.runAgent, and agent.subscribe primitives above still apply. Only the interrupt path depends on the framework.

For every method on the core, see the useCopilotKit reference.

How is this guide?