Menu
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/ag2ai/resource-hub --skill build-group-chatThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
REST and WebSocket endpoint patterns, error handling, and Pydantic schema conventions for the backend
Architecture, directory layout, communication protocol, and conventions for the full-stack multi-agent application
Step-by-step guide to add a new REST or WebSocket endpoint to the backend
Showing SKILL.md
Based on SOC occupation classification
| name | build-group-chat |
| description | Step-by-step workflow for building an AG2 multi-agent group chat |
| license | Apache-2.0 |
from ag2 import LLMConfig
from ag2.agentchat import AssistantAgent, UserProxyAgent
from ag2.agentchat.group import run_group_chat, DefaultPattern, Handoff
Define agents inside an LLMConfig context manager. Give each agent a distinct role.
with LLMConfig(api_type="openai", model="gpt-4o"):
planner = AssistantAgent(
name="planner",
system_message=(
"You break down user requests into actionable steps. "
"Hand off to coder when the plan is ready."
),
)
coder = AssistantAgent(
name="coder",
system_message=(
"You write Python code based on the plan. "
"Hand off to reviewer when code is complete."
),
)
reviewer = AssistantAgent(
name="reviewer",
system_message=(
"You review code for correctness and style. "
),
)
executor = UserProxyAgent(
name=,
human_input_mode=,
)
Agents explicitly hand off to the next speaker. You define allowed transitions.
pattern = DefaultPattern(
initial_agent=planner,
agents=[planner, coder, reviewer, executor],
handoffs=[
Handoff(source=planner, target=coder),
Handoff(source=coder, target=reviewer),
Handoff(source=reviewer, target=coder),
],
group_manager_args={"llm_config": LLMConfig(api_type="openai", model="gpt-4o")},
)
The LLM decides who speaks next. No explicit handoffs needed.
from ag2.agentchat.group import AutoPattern
pattern = AutoPattern(
initial_agent=planner,
agents=[planner, coder, reviewer, executor],
group_manager_args={"llm_config": LLMConfig(api_type="openai", model="gpt-4o")},
)
Agents speak in the order they are listed. Simple and predictable.
from ag2.agentchat.group import RoundRobinPattern
pattern = RoundRobinPattern(
initial_agent=planner,
agents=[planner, coder, reviewer],
)
result = run_group_chat(
pattern=pattern,
messages="Build a CLI tool that converts CSV files to JSON.",
)
from ag2 import LLMConfig
from ag2.agentchat import AssistantAgent, UserProxyAgent
from ag2.agentchat.group import run_group_chat, DefaultPattern, Handoff
with LLMConfig(api_type="openai", model="gpt-4o"):
planner = AssistantAgent(
name="planner",
system_message="You create step-by-step plans. Hand off to coder when ready.",
)
coder = AssistantAgent(
name="coder",
system_message="You implement code from plans. Hand off to reviewer when done.",
)
reviewer = AssistantAgent(
name="reviewer",
system_message=(
"You review code quality. Hand off to coder for fixes. "
"Reply TERMINATE when approved."
),
)
executor = UserProxyAgent(name="executor", human_input_mode="NEVER")
pattern = DefaultPattern(
initial_agent=planner,
agents=[planner, coder, reviewer, executor],
handoffs=[
Handoff(source=planner, target=coder),
Handoff(source=coder, target=reviewer),
Handoff(source=reviewer, target=coder),
],
group_manager_args={"llm_config": LLMConfig(api_type="openai", model="gpt-4o")},
)
result = run_group_chat(
pattern=pattern,
messages="Build a function that merges two sorted lists.",
)