Skip to content

Quick Start

This guide will get you up and running with a simple AI agent in minutes.

1. Setup API Keys

Create a .env file in your project root and add your API keys:

OPENAI_API_KEY=sk-...
GEMINI_API_KEY=AIza...
ANTHROPIC_API_KEY=sk-ant...

2. Create Your First Agent

Here is a simple example using OpenRouter (which supports many models). You can replace OpenRouterClient with OpenAIClient, GeminiClient, etc.

import os
from dotenv import load_dotenv
from agent_sdk import Runner, Agent, OpenRouterClient

# 1. Load Environment Variables
load_dotenv()

# 2. Initialize Client and Runner
client = OpenRouterClient(api_key=os.getenv("OPENRouter_API_KEY"))
runner = Runner(client)

# 3. Define the Agent
assistant = Agent(
    name="Assistant",
    model="mistralai/mistral-7b-instruct",
    instructions="You are a helpful assistant who answers concisely."
)

# 4. Run the Agent (Streaming)
print(f"Chatting with {assistant.name}...
")
stream = runner.run_stream(assistant, "Hello! Who are you?")

for event in stream:
    if event.type == "token":
        print(event.data, end="", flush=True)
    elif event.type == "final":
        print("\n\n[Done]")

3. What Just Happened?

  1. Client: We initialized a client to communicate with the LLM provider.
  2. Runner: The Runner class manages the conversation loop, tool execution, and memory.
  3. Agent: We defined an Agent with a name, a specific model, and system instructions.
  4. Streaming: runner.run_stream returns a generator that yields events (tokens, tool calls, errors) in real-time.

Next Steps

  • Learn about Universal Clients to use different providers.
  • Explore Agents & Runners for more configuration options.
  • Add Tools to give your agent capabilities like web search or file access.