v2.5

Teams now support four distinct execution modes (coordinate, route, broadcast, and tasks), giving you explicit control over how agents collaborate. Instead of building custom orchestration logic, you select the mode that matches your use case: route requests to a specialist, coordinate across multiple agents, broadcast to all members in parallel, or decompose work into discrete tasks.

This reduces orchestration boilerplate, makes team behavior predictable and auditable, and lets you switch strategies without rewriting agent logic.

Details

• Coordinate mode lets a lead agent delegate and synthesize across team members
• Route mode directs each request to the best-fit agent based on the task
• Broadcast sends the same input to all members and collects parallel responses
• Tasks mode decomposes work into discrete, trackable units assigned to individual agents
• Configured via the TeamMode parameter when defining a team

‍

from agno.agent import Agent
from agno.team.mode import TeamMode
from agno.team.team import Team
from agno.models.openai import OpenAIChat

# --- Specialist Agents (Team Members) ---

researcher = Agent(
    name="Researcher",
    model=OpenAIChat(id="gpt-4o"),
    instructions="You are a research specialist. Find and summarize factual information on any given topic.",
)

analyst = Agent(
    name="Analyst",
    model=OpenAIChat(id="gpt-4o"),
    instructions="You are a data analyst. Interpret findings and extract key insights from research.",
)

writer = Agent(
    name="Writer",
    model=OpenAIChat(id="gpt-4o"),
    instructions="You are a technical writer. Turn insights into clear, concise reports.",
)

# --- Team with Coordinate Mode ---
# Swap TeamMode.coordinate for .route, .broadcast, or .tasks as needed

team = Team(
    name="Research Team",
    mode=TeamMode.coordinate,  # Leader delegates and synthesizes across all members
    members=[researcher, analyst, writer],
    model=OpenAIChat(id="gpt-4o"),  # Model used by the team leader
    instructions="Coordinate the specialist agents to produce a well-researched, clearly written report.",
)

# --- Run the Team ---
response = team.print_response("Give me a report on the current state of quantum computing.", stream=True, show_member_responses=True)

‍

View the docs