Add Agent To Team

Add an Agent to the Team

Steps

1. Define the agent in backend/app/agents.py

Add the agent inside the with config: block alongside the existing agents:

with config:
    planner = AssistantAgent(...)
    coder = AssistantAgent(...)
    reviewer = AssistantAgent(...)

    # New agent
    tester = AssistantAgent(
        name="tester",
        system_message=(
            "You are a testing agent. Given the coder's implementation, "
            "write comprehensive unit tests covering edge cases and error "
            "conditions. Use pytest conventions."
        ),
    )

2. Add the agent to the AGENTS list

AGENTS = [planner, coder, reviewer, tester]

AGENT_NAMES updates automatically since it's derived from AGENTS.

3. Update the pattern if needed

For RoundRobinPattern, agents execute in list order — place the new agent where it should speak in the sequence.

If switching to AutoPattern for dynamic routing:

from autogen.agentchat.group.patterns.pattern import AutoPattern

pattern = AutoPattern(
    initial_agent=planner,
    agents=AGENTS,
)

4. Update max_rounds in run_team()

Each agent takes one round. If you added an agent, increase max_rounds to allow at least 2 full cycles:

result = await run_group_chat(
    pattern=pattern,
    messages=message,
    max_rounds=8,  # was 6 with 3 agents
)

5. Add the agent color in the frontend

In frontend/src/components/Chat.tsx, add a color entry:

const AGENT_COLORS: Record<string, string> = {
  planner: "#2563eb",
  coder: "#059669",
  reviewer: "#d97706",
  tester: "#7c3aed",   // new
  system: "#6b7280",
};

6. Update the test

In backend/tests/test_api.py:

async def test_health(client: AsyncClient) -> None:
    resp = await client.get("/api/health")
    body = resp.json()
    assert "tester" in body["agents"]

Checklist

• Agent defined with a specific, focused system message
• Added to AGENTS list in the correct position
• max_rounds updated to accommodate the new agent
• Frontend color mapping added
• Health check test updated