TeamFactory

Build a Team per request whose members, model, and instructions depend on the caller.

v2.6.0

A TeamFactory produces a fresh Team for each request. Register it in AgentOS(teams=[...]) alongside any prototype teams.

basic_team_factory.py

from agno.agent import Agent
from agno.db.postgres import PostgresDb
from agno.factory import RequestContext
from agno.models.openai import OpenAIResponses
from agno.os import AgentOS
from agno.team import Team, TeamFactory

db = PostgresDb(db_url="postgresql+psycopg://ai:ai@localhost:5532/ai")


def build_support_team(ctx: RequestContext) -> Team:
    user_id = ctx.user_id or "anonymous"

    billing_agent = Agent(
        name="Billing Agent",
        role="Handle billing inquiries",
        model=OpenAIResponses(id="gpt-5.4"),
        instructions=f"You handle billing questions for tenant {user_id}. Be concise.",
    )

    tech_agent = Agent(
        name="Tech Support Agent",
        role="Handle technical issues",
        model=OpenAIResponses(id="gpt-5.4"),
        instructions=f"You handle technical support for tenant {user_id}. Be concise.",
    )

    return Team(
        name="Support Team",
        model=OpenAIResponses(id="gpt-5.4"),
        members=[billing_agent, tech_agent],
        db=db,
        instructions=[
            f"You are the support team leader for tenant {user_id}.",
            "Route billing questions to the Billing Agent and technical issues to the Tech Support Agent.",
        ],
        markdown=True,
    )


support_team_factory = TeamFactory(
    id="support-team",
    db=db,
    factory=build_support_team,
    name="Per-tenant Support Team",
    description="Builds a support team with billing and tech agents per tenant.",
)

agent_os = AgentOS(teams=[support_team_factory])
app = agent_os.get_app()

if __name__ == "__main__":
    agent_os.serve(app="basic_team_factory:app", port=7777, reload=True)

Run the team like any other team:

curl -X POST http://localhost:7777/teams/support-team/runs \
    -F 'message=I need help with billing and a technical issue' \
    -F 'user_id=tenant_42' \
    -F 'stream=false'

See the Factories reference for the full constructor signature and parameter list.

Members That Scale with the Caller

Member composition can change per request. A common pattern: scale team size and model quality with a subscription tier from ctx.trusted.claims.

def build_research_team(ctx: RequestContext) -> Team:
    tier = ctx.trusted.claims.get("tier", "free")
    model_id = TIER_MODELS.get(tier, "gpt-4.1-mini")

    members = [researcher(model_id), writer(model_id)]
    if tier == "enterprise":
        members.append(reviewer(model_id))

    return Team(name="Research Team", model=OpenAIResponses(id=model_id), members=members, db=db)

The tier claim comes from verified middleware, so the tier cannot be changed by the request body. See Authorization From Verified Context for the trust model and JWT Role Factory for an end-to-end JWT example.

Async Factories

Use an async callable when you need to fetch context from a database, an HTTP service, or any other awaitable resource.

async_team_factory.py

async def build_team(ctx: RequestContext) -> Team:
    members = await load_members_for_tenant(ctx.user_id)
    return Team(
        name="Tenant Team",
        model=OpenAIResponses(id="gpt-5.4"),
        members=members,
        db=db,
    )


team_factory = TeamFactory(id="tenant-team", db=db, factory=build_team)

With an Input Schema

Declare a Pydantic model on the factory to validate client-supplied factory_input before the factory runs.

team_factory_input.py

from pydantic import BaseModel


class TeamConfig(BaseModel):
    include_reviewer: bool = False


def build_team(ctx: RequestContext) -> Team:
    cfg: TeamConfig = ctx.input
    members = [researcher, writer]
    if cfg.include_reviewer:
        members.append(reviewer)
    return Team(name="Research", model=OpenAIResponses(id="gpt-5.4"), members=members, db=db)


team_factory = TeamFactory(
    id="research-team",
    db=db,
    factory=build_team,
    input_schema=TeamConfig,
)

AgentFactory and TeamFactory instances cannot be passed inside members. Build per-member customization inside the team’s factory.

Error Handling

Raise FactoryPermissionError from inside the factory to reject unauthorized callers with HTTP 403. AgentOS raises FactoryValidationError (400) automatically when factory_input fails input_schema validation.

from agno.factory import FactoryPermissionError


def build_team(ctx: RequestContext) -> Team:
    if "teams:run" not in ctx.trusted.scopes:
        raise FactoryPermissionError("Missing 'teams:run' scope")
    ...

See the Factories reference for the full exception hierarchy and the post-resolve behavior.

Developer Resources

• Factories overview
• RequestContext fields
• Factories reference
• Factory examples