Skip to main content
AutoGen models multi-agent collaboration as a conversation — agents exchange messages until a termination condition is met. The Swarms API maps this same collaborative pattern onto structured workflows with a clean REST interface, eliminating the need to manage conversation loops, termination strings, and local LLM configuration.
AutoGenSwarms API
AssistantAgent(name, system_message, llm_config)agent with agent_name, system_prompt, model_name
UserProxyAgentNot needed; task input is the top-level task field
GroupChat(agents, messages, max_round)GroupChat swarm type
GroupChatManager(groupchat, llm_config)Managed by the API
agent.initiate_chat(recipient, message)POST request with task field
ConversableAgentAny agent with system_prompt
llm_config = {"model": "gpt-4o", "api_key": ...}"model_name": "gpt-4o" on agent spec
max_consecutive_auto_replymax_loops on agent spec
is_termination_msgHandled by workflow max_loops
code_execution_config"tools": ["code_interpreter"] on agent spec

​
Side-by-Side: Basic Two-Agent Chat

​
AutoGen

import autogen

config_list = [{"model": "gpt-4o", "api_key": "sk-..."}]

assistant = autogen.AssistantAgent(
    name="assistant",
    system_message="You are a helpful AI assistant.",
    llm_config={"config_list": config_list},
)

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=3,
    is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE"),
    code_execution_config=False,
)

user_proxy.initiate_chat(
    assistant,
    message="Explain the difference between supervised and unsupervised learning.",
)

​
Swarms API

import os
import requests

result = requests.post(
    "https://api.swarms.world/v1/agent/completions",
    headers={"x-api-key": os.environ["SWARMS_API_KEY"], "Content-Type": "application/json"},
    json={
        "agent_name": "assistant",
        "system_prompt": "You are a helpful AI assistant.",
        "task": "Explain the difference between supervised and unsupervised learning.",
        "model_name": "gpt-4o",
        "max_loops": 1,
        "temperature": 0.5,
    },
    timeout=60,
).json()

print(result["outputs"])
For a simple two-agent conversation where UserProxy only relays a message and Assistant responds, a single agent completion is all you need. The UserProxyAgent is not an AI agent — it just passes the task.

​
Side-by-Side: GroupChat

​
AutoGen

import autogen

config_list = [{"model": "gpt-4o", "api_key": "sk-..."}]
llm_config = {"config_list": config_list, "cache_seed": 42}

coder = autogen.AssistantAgent(
    name="Coder",
    system_message="You are an expert software engineer. Write clean, efficient code.",
    llm_config=llm_config,
)

reviewer = autogen.AssistantAgent(
    name="Code_Reviewer",
    system_message="You are a senior code reviewer. Review code for bugs and best practices.",
    llm_config=llm_config,
)

product_manager = autogen.AssistantAgent(
    name="Product_Manager",
    system_message="You are a product manager. Ensure solutions meet business requirements.",
    llm_config=llm_config,
)

user_proxy = autogen.UserProxyAgent(
    name="User_Proxy",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=0,
    code_execution_config=False,
)

groupchat = autogen.GroupChat(
    agents=[user_proxy, coder, reviewer, product_manager],
    messages=[],
    max_round=5,
)
manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=llm_config)

user_proxy.initiate_chat(
    manager,
    message="Build a Python function that rates the readability of a text on a 1-10 scale.",
)

​
Swarms API

import os
import requests

result = requests.post(
    "https://api.swarms.world/v1/swarms/completions",
    headers={"x-api-key": os.environ["SWARMS_API_KEY"], "Content-Type": "application/json"},
    json={
        "name": "Software Development GroupChat",
        "description": "Collaborative group: coder, reviewer, and PM",
        "swarm_type": "GroupChat",
        "task": "Build a Python function that rates the readability of a text on a 1-10 scale.",
        "agents": [
            {
                "agent_name": "Coder",
                "system_prompt": "You are an expert software engineer. Write clean, efficient, well-documented code.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.3,
            },
            {
                "agent_name": "Code_Reviewer",
                "system_prompt": "You are a senior code reviewer. Review code for bugs, edge cases, and best practices. Provide specific, actionable feedback.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.2,
            },
            {
                "agent_name": "Product_Manager",
                "system_prompt": "You are a product manager. Ensure solutions meet business requirements, are maintainable, and deliver real user value.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.3,
            },
        ],
        "max_loops": 5,
    },
    timeout=180,
).json()

print(result["outputs"])
What changed:
  • UserProxyAgent → removed; the task field replaces it
  • GroupChatManager → managed by the API
  • max_round=5 → "max_loops": 5 at the top level
  • cache_seed → not needed; the API is stateless
  • llm_config per agent → "model_name" per agent

​
Side-by-Side: Code Execution Agent

​
AutoGen

import autogen

config_list = [{"model": "gpt-4o", "api_key": "sk-..."}]

assistant = autogen.AssistantAgent(
    name="assistant",
    llm_config={"config_list": config_list},
)

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    code_execution_config={
        "work_dir": "coding",
        "use_docker": False,
    },
)

user_proxy.initiate_chat(
    assistant,
    message="Write and execute a Python script that calculates the first 20 Fibonacci numbers.",
)

​
Swarms API

import os
import requests

result = requests.post(
    "https://api.swarms.world/v1/agent/completions",
    headers={"x-api-key": os.environ["SWARMS_API_KEY"], "Content-Type": "application/json"},
    json={
        "agent_name": "coding_assistant",
        "system_prompt": "You are an expert programmer. Write correct, efficient Python code and explain what it does.",
        "task": "Write a Python script that calculates the first 20 Fibonacci numbers.",
        "model_name": "gpt-4o",
        "tools": ["code_interpreter"],
        "max_loops": 1,
        "temperature": 0.2,
    },
    timeout=90,
).json()

print(result["outputs"])

​
Side-by-Side: Multi-Agent Debate Pattern

AutoGen is often used for debate-style workflows where agents argue positions. The Swarms API has a dedicated DebateWithJudge architecture for this.

​
AutoGen

import autogen

config_list = [{"model": "gpt-4o", "api_key": "sk-..."}]

pro_agent = autogen.AssistantAgent(
    name="Pro_AI",
    system_message="You argue FOR AI replacing human jobs. Present strong evidence.",
    llm_config={"config_list": config_list},
)

con_agent = autogen.AssistantAgent(
    name="Con_AI",
    system_message="You argue AGAINST AI replacing human jobs. Present strong evidence.",
    llm_config={"config_list": config_list},
)

judge = autogen.AssistantAgent(
    name="Judge",
    system_message="You are a neutral judge. After both sides present, give a balanced verdict.",
    llm_config={"config_list": config_list},
)

groupchat = autogen.GroupChat(
    agents=[pro_agent, con_agent, judge],
    messages=[],
    max_round=6,
)
manager = autogen.GroupChatManager(groupchat=groupchat, llm_config={"config_list": config_list})
pro_agent.initiate_chat(manager, message="Will AI replace most human jobs in the next 20 years?")

​
Swarms API

import os
import requests

result = requests.post(
    "https://api.swarms.world/v1/swarms/completions",
    headers={"x-api-key": os.environ["SWARMS_API_KEY"], "Content-Type": "application/json"},
    json={
        "name": "AI Jobs Debate",
        "description": "Two agents debate, a judge decides",
        "swarm_type": "DebateWithJudge",
        "task": "Will AI replace most human jobs in the next 20 years?",
        "agents": [
            {
                "agent_name": "Pro_AI",
                "system_prompt": "You argue FOR AI replacing most human jobs in the next 20 years. Present compelling evidence, statistics, and historical precedents.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.6,
            },
            {
                "agent_name": "Con_AI",
                "system_prompt": "You argue AGAINST AI replacing most human jobs. Present economic theory, counterexamples, and evidence for human adaptability.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.6,
            },
            {
                "agent_name": "Judge",
                "system_prompt": "You are a neutral judge and critical thinker. After both sides present their arguments, evaluate the quality of evidence, logical coherence, and give a balanced, reasoned verdict.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.3,
            },
        ],
        "max_loops": 3,
    },
    timeout=180,
).json()

print(result["outputs"])

​
LLM Configuration Migration

AutoGen requires per-agent llm_config dicts with API keys and model lists. The Swarms API handles all authentication — you just specify the model name.

​
AutoGen

config_list = [
    {
        "model": "gpt-4o",
        "api_key": os.environ["OPENAI_API_KEY"],
        "base_url": "https://api.openai.com/v1",
    }
]

llm_config = {
    "config_list": config_list,
    "temperature": 0.7,
    "max_tokens": 2000,
    "cache_seed": 42,
    "timeout": 120,
}

agent = autogen.AssistantAgent(
    name="my_agent",
    llm_config=llm_config,
)

​
Swarms API

# No API keys per model, no config_list, no cache_seed
{
    "agent_name": "my_agent",
    "system_prompt": "...",
    "model_name": "gpt-4o",
    "temperature": 0.7,
    "max_tokens": 2000,
    "max_loops": 1,
}
The Swarms API supports 300+ models. See Available Models for the full list.

​
Termination Conditions

AutoGen uses is_termination_msg callbacks to stop conversations. The Swarms API uses max_loops to bound execution.

​
AutoGen

def termination_check(msg):
    return msg.get("content", "").rstrip().endswith("TERMINATE")

agent = autogen.UserProxyAgent(
    name="user",
    is_termination_msg=termination_check,
    max_consecutive_auto_reply=10,
)

​
Swarms API

{
    "swarm_type": "GroupChat",
    "task": "...",
    "agents": [...],
    "max_loops": 5,
}

​
Key Differences to Keep in Mind

ConcernAutoGenSwarms API
Conversation historyManaged in-memory per sessionStateless; each request is independent
Human inputhuman_input_mode="ALWAYS"Not supported in API mode
Code executionDocker or local subprocessCloud sandbox via "tools": ["code_interpreter"]
Function callingregister_function()"tools" array with built-in tool names
Nested chatsregister_nested_chats()Use HierarchicalSwarm or GraphWorkflow
Cost / token trackingVia OpenAI usage logsusage.token_cost in every response
Cachingcache_seed on llm_configNot applicable; no local cache