Use codemap find "SymbolName" to locate any symbol before reading — it gives exact file + line range and is faster than grep across the whole repo.

Building Multi-Agent Teams

Quick start

Define a team and run it:

from __future__ import annotations

from lauren_ai import team, TeamRunner
from .agents import ResearchAgent, WriterAgent

@team(
    name="content-team",
    mode="collaborate",
    model="claude-opus-4-6",
)
class ContentTeam:
    def __init__(self, researcher: ResearchAgent, writer: WriterAgent) -> None:
        self.researcher = researcher
        self.writer = writer


# Run
runner = TeamRunner(
    team_cls=ContentTeam,
    llm=llm_service,
    agent_runner=agent_runner,
)
result = await runner.run("Write a report on quantum computing.")
print(result.final_answer)

Rules:

• @team() must use parentheses.
• Worker parameters in __init__ must be type-annotated with their agent class.
• TeamRunner discovers workers by reading __init__.__annotations__.

Modes

Streaming events

TeamRunner.run_stream() yields TeamEvent subclass instances:

from lauren_ai import (
    TeamWorkerStarted,
    TeamWorkerFinished,
    TeamCoordinatorDecision,
    TeamFinalAnswer,
)

async for event in runner.run_stream("Research quantum computing"):
    if isinstance(event, TeamWorkerStarted):
        print(f"[Round {event.round}] Starting: {event.worker_name}")
    elif isinstance(event, TeamWorkerFinished):
        print(f"[Round {event.round}] {event.worker_name} done")
    elif isinstance(event, TeamCoordinatorDecision):
        print(f"[Coordinator] {event.decision}")
    elif isinstance(event, TeamFinalAnswer):
        print(f"[Final] {event.content}")

TeamResult fields

Reference

See teams.md for:

• Full @team() parameter reference (coordinator_prompt, max_rounds)
• Custom coordinator prompt with {worker_descriptions}, {task}, {prior_outputs} placeholders
• SSE streaming in a web controller with EventStream and ServerSentEvent
• collaborate vs coordinator mode detailed behaviour