Documentation

AutoGen Integration

Track Microsoft AutoGen multi-agent conversations with automatic message tracking and tool executions.

What Gets Tracked Automatically

Agent Messages

All agent-to-agent communications.

Tool Calls

Function and tool executions.

Model Responses

LLM responses with token usage.

Cost & Tokens

Estimated costs based on model.

Installation

# For AutoGen 0.4+ (recommended)
pip install guardy autogen-agentchat>=0.4.0

# For legacy pyautogen
pip install guardy pyautogen>=0.2.0

AutoGen 0.4+ (autogen-agentchat)

from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.teams import RoundRobinGroupChat
from autogen_ext.models.openai import OpenAIChatCompletionClient
from guardy import GuardyClient
from guardy.autogen import GuardyAutogenHandler

# Initialize
client = GuardyClient(api_key="guardy_live_xxx")
session_id = client.create_session(
    name="Code Review Team",
    agent_name="autogen-review",
    user_id="user_123"
)

# Create handler
handler = GuardyAutogenHandler(client, session_id, verbose=True)

# Create model client
model_client = OpenAIChatCompletionClient(model="gpt-4o")

# Create agents
reviewer = AssistantAgent(
    name="CodeReviewer",
    model_client=model_client,
    system_message="You review code for bugs.",
)

security = AssistantAgent(
    name="SecurityExpert",
    model_client=model_client,
    system_message="You analyze security vulnerabilities.",
)

# Create team
team = RoundRobinGroupChat(
    participants=[reviewer, security],
    max_turns=4,
)

# Run with tracking
result = await handler.run_team(team, task="Review this code...")

# Finish
handler.finish(success=True)

Legacy pyautogen

from autogen import AssistantAgent, UserProxyAgent
from guardy import GuardyClient
from guardy.autogen import GuardyAutogenHandler

client = GuardyClient(api_key="guardy_live_xxx")
session_id = client.create_session(
    name="AutoGen Chat",
    agent_name="autogen-chat",
    user_id="user_123"
)

handler = GuardyAutogenHandler(client, session_id)

# Create agents
assistant = AssistantAgent("assistant", llm_config=llm_config)
user_proxy = UserProxyAgent("user_proxy")

# Register for tracking
handler.register_agents([assistant, user_proxy])

# Chat
user_proxy.initiate_chat(assistant, message="Hello!")

handler.finish(success=True)

Next Steps

Python SDK Reference

Full SDK documentation.

CrewAI Integration

Track CrewAI multi-agent systems.