Skip to main content
This guide shows how to combine Nango with any LLM provider to consume external APIs via tool calling.

​
Example overview

The example below uses Hubspot’s API to demonstrate how to access a user’s external account through Nango.
Requirements
  • A Nango account with an integration for hubspot and the whoami action enabled.
  • An account with your preferred LLM provider.
In this example:
  1. Check if the user is already authenticated with Hubspot.
  2. If not, start the Hubspot OAuth flow. You can do this:
    • Directly in the chat session, or
    • By embedding the Frontend SDK in your app.
  3. Send a prompt to your model.
  4. When the model calls a tool, trigger the corresponding Nango action.

​
Example code

import { Nango } from "@nangohq/node";

// Initialize the Nango client
const nango = new Nango({ secretKey: process.env.NANGO_SECRET_KEY! });

export async function runLLMAgent(modelClient: any) {
  const userId = "user1";
  const integrationId = "hubspot";

  // Step 1: Ensure the user is authorized
  console.log("🔒 Checking Hubspot authorization...");
  let connectionId = (await nango.listConnections({ userId })).connections[0]?.connection_id;

  // Step 2: If the user is not authorized, redirect to the auth flow.
  if (!connectionId) {
    console.log("User not authorized, starting Hubspot auth flow...");

    const session = await nango.createConnectSession({
      allowed_integrations: [integrationId],
      end_user: { id: userId },
    }); // Handle API auth via Nango.

    // Redirect the user to a page to authorize Hubspot. 
    // Alternatively, you can embed Nango's Frontend SDK in your app.
    console.log(`Please authorize Hubspot here: ${session.data.connect_link}`);

    const connection = await nango.waitForConnection(integrationId, userId);
    connectionId = connection?.connection_id;

    if (!connectionId) throw new Error("Auth flow failed");
  }

  // Step 3: Send a prompt to your model.
  console.log("🤖 Agent running...");
  const response = await modelClient.generate({
    model: "your-llm-model-id",
    input: "Get the current Hubspot user info using the who_am_i tool.",
    tools: [
      {
        name: "who_am_i",
        description: "Fetch the current Hubspot user profile.",
        parameters: { type: "object", properties: {} },
      },
    ],
  });

  // Check if the model requested a tool call
  const toolCall = response?.toolCalls?.[0] || response?.tool_call || null;

  if (toolCall?.name === "who_am_i") {
    console.log("🧩 Model decided to call the 'who_am_i' tool...");

    try {
      // Step 4: Execute the requested tool with Nango.
      const userInfo = await nango.triggerAction("hubspot", connectionId, "whoami");
      console.log("✅ Retrieved Hubspot user info:", userInfo);
    } catch (error: any) {
      console.error("❌ Nango API error:", error.response?.data?.error || error);
    }
  } else {
    // Handle plain text or other model responses
    const textOutput = response?.text || response?.output_text || JSON.stringify(response);
    console.log("🗣️ Model response:", textOutput);
  }
}
This example uses the Nango Node SDK, which is a thin wrapper around Nango’s REST API. You can use the same flow in any language with the API reference.
You can let agents or models introspect available tools and their specifications by calling this
endpoint. Your model can then automatically decide which tool to call.