​
Quick Start: Add a Backend to Your Agent

This guide provides a simple Hello World example to get you started with xpander’s backend services.You’ll learn:
  • How to connect to xpander’s backend services
  • Basic agent execution flow
  • How to process tool calls
  • How to manage conversation state
  • Core SDK architecture and components
This guide shows how to enhance any agent with xpander.ai’s backend for memory, tools, and more.

​
📦 Installation

# Python
pip install xpander-sdk

# Node.js
npm install @xpander-ai/sdk

# CLI (for agent creation)
npm install -g xpander-cli

​
Use xpander-cli to scaffold a new agent template

xpander login
xpander agent new
python xpander_handler.py  # Entry point for your agent's event handling

​
Create Hello World Agent

hello_world_agent.py
# First, install the SDK if you haven't already
# pip install xpander-sdk

from xpander_sdk import XpanderClient, LLMProvider
from openai import OpenAI
import os

# Initialize clients with your API keys
# Set these environment variables before running
xpander = XpanderClient(api_key=os.getenv("XPANDER_API_KEY"))
openai = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

# Get your agent from xpander's backend
# Create your agent first at https://app.xpander.ai
agent = xpander.agents.get(agent_id=os.getenv("XPANDER_AGENT_ID"))

# Select the LLM provider you want to use
agent.select_llm_provider(LLMProvider.OPEN_AI)

# Let the LLM decide when it has finished
agent.enable_agent_end_tool()

# Add a task for the agent to process
agent.add_task("Hello world! Tell me what you can do.")

# Run the agent with OpenAI
while not agent.is_finished():
    # Get LLM response from OpenAI
    response = openai.chat.completions.create(
        model="gpt-4.1",
        messages=agent.messages,
        tools=agent.get_tools(),
        tool_choice=agent.tool_choice,
        temperature=0
    )
    
    # Add response to agent memory
    agent.add_messages(response.model_dump())
    
    # Extract and run any tools the agent wants to use
    tool_calls = agent.extract_tool_calls(
        llm_response=response.model_dump()
    )
    if tool_calls:
        agent.run_tools(tool_calls=tool_calls)

# Get the final result
result = agent.retrieve_execution_result()
print(result.result)

# Store the thread_id for future conversations (memory persistence)
thread_id = result.memory_thread_id
print(f"Thread ID: {thread_id}")
hello_world_agent.ts
import { XpanderClient, LLMProvider } from '@xpander-ai/sdk';
import OpenAI from 'openai';

const xpander = new XpanderClient({ apiKey: process.env.XPANDER_API_KEY });
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

const agent = await xpander.agents.get({ agentId: process.env.XPANDER_AGENT_ID });
agent.selectLLMProvider(LLMProvider.OPEN_AI);
// Allow the model to finish when done
agent.enableAgentEndTool();
await agent.addTask({ input: 'Hello world! Tell me what you can do.' });

while (!(await agent.isFinished())) {
  const response = await openai.chat.completions.create({
    model: 'gpt-4.1',
    messages: agent.messages,
    tools: agent.getTools(),
    tool_choice: agent.toolChoice,
    temperature: 0
  });

  await agent.addMessages(response);
  const toolCalls = agent.extractToolCalls({ llmResponse: response });
  if (toolCalls.length > 0) {
    await agent.runTools({ toolCalls });
  }
}

const result = await agent.retrieveExecutionResult();
console.log(result.result);
const threadId = result.memoryThreadId;
console.log(`Thread ID: ${threadId}`);

​
Deploy agent to the cloud

xpander deploy  # Deploys the Docker container to the cloud and runs it via xpander_handler.py
xpander logs    # Streams logs locally from your configured agent

​
What This Does

The example above:
  1. Connects to xpander.ai’s backend services
  2. Loads an agent you’ve created in the Builder
  3. Adds a task for the agent to process
  4. Runs the agent with the OpenAI API
  5. Executes any tools the agent needs
  6. Returns the result with a thread ID for continuity
See the full Hello World application here.

​
Using Different LLM Providers

The xpander.ai backend is model-agnostic. You can use any LLM provider: 
from openai import OpenAI

openai_client = OpenAI(api_key="OpenAI-APIkey")

# First, select the provider
agent.select_llm_provider(LLMProvider.OPEN_AI)

# Then use the provider's client
response = openai_client.chat.completions.create(
    model="gpt-4.1",
    messages=agent.messages,
    tools=agent.get_tools(),  # Tools automatically formatted for OpenAI
    tool_choice=agent.tool_choice
)

# Process the response
agent.add_messages(response.model_dump())
tool_calls = agent.extract_tool_calls(
    llm_response=response.model_dump()
)
agent.run_tools(tool_calls=tool_calls)