Setup

Three steps or fewer.

STEP 01

Install

pip install autogen-agentchat autogen-ext[openai]

STEP 02

Configure the model client

OpenAIChatCompletionClient accepts base_url and api_key. Use AIgateway's endpoint and key — every agent that takes this client routes through us.

from autogen_ext.models.openai import OpenAIChatCompletionClient

client = OpenAIChatCompletionClient(
    model="anthropic/claude-opus-4.7",
    base_url="https://api.aigateway.sh/v1",
    api_key="sk-aig-...",
)

STEP 03

Use across an agent graph

Hand the client to AssistantAgent, UserProxyAgent, RoundRobinGroupChat — wherever AutoGen expects a model. Tool calls, structured output, and streaming all work.

from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.teams import RoundRobinGroupChat
from autogen_agentchat.conditions import MaxMessageTermination

writer = AssistantAgent("writer", client, system_message="Draft.")
critic = AssistantAgent("critic", client, system_message="Critique.")
team = RoundRobinGroupChat(
    [writer, critic],
    termination_condition=MaxMessageTermination(6),
)
await team.run(task="Write a paragraph on edge inference.")

Notes

AutoGen v0.4 ships with native model_info passing — make sure model_info matches the AIgateway model's capabilities (tool calling, vision, json mode) for the right client behavior.
Different agents can use different AIgateway models — just instantiate multiple OpenAIChatCompletionClients with different model strings.
Pair with /v1/evals to grade an AutoGen pipeline across candidate models on real conversation traces.

AutoGen + AIgateway

Three steps or fewer.

Install

Configure the model client

Use across an agent graph

Same key. Every other tool.