Skip to main content

Getting Started

This guide walks you through installing Agent Teams, creating your first team, and running tasks.

Prerequisites​

  • Python 3.10+
  • A model provider API key (OpenAI, Anthropic, etc.)

Installation​

pip install agent-teams

For development:

git clone https://github.com/datalayer/agent-teams
cd agent-teams
pip install -e ".[test]"

Option A: Python API​

1. Create a Team​

import asyncio
from agent_teams.types import TeamConfig, AgentMemberConfig, ExecutionMode
from agent_teams.manager import TeamManager

async def main():
    manager = TeamManager()

    config = TeamConfig(
        name="my-first-team",
        execution_mode=ExecutionMode.SEQUENTIAL,
        members=[
            AgentMemberConfig(
                name="analyst",
                model="openai:gpt-4o",
                system_prompt="You are a data analyst.",
            ),
        ],
    )
    team = await manager.create_team(config)
    print(f"Team created: {team.team_id}")

asyncio.run(main())

2. Start the Team and Assign a Task​

await manager.start_team(team.team_id)

result = await manager.assign_task(
    team.team_id,
    "Summarise the key trends in Q4 revenue data",
)
print(result)

3. Check Metrics​

metrics = manager.get_metrics(team.team_id)
print(f"Tasks completed: {metrics.tasks_completed}")
print(f"Tasks failed: {metrics.tasks_failed}")

4. Stop the Team​

await manager.stop_team(team.team_id)

Option B: CLI​

1. Start the Server​

agent-teams serve --port 8765

2. Create a Team from a YAML File​

Create a team.yaml file:

name: my-first-team
execution_mode: sequential
members:
  - name: analyst
    model: openai:gpt-4o
    system_prompt: You are a data analyst.

Then create the team:

agent-teams create team.yaml

3. Manage the Team​

# List all teams
agent-teams list

# Start the team
agent-teams start <team-id>

# Assign a task
agent-teams assign <team-id> "Summarise Q4 revenue trends"

# Check status
agent-teams status <team-id>

# View metrics
agent-teams metrics <team-id>

# Stream events
agent-teams events <team-id>

# Stop the team
agent-teams stop <team-id>

Option C: REST API​

The FastAPI server exposes a full REST interface:

# Create a team
curl -X POST http://localhost:8765/teams \
  -H "Content-Type: application/json" \
  -d '{"name": "my-team", "execution_mode": "sequential", "members": [{"name": "analyst", "model": "openai:gpt-4o"}]}'

# Start the team
curl -X POST http://localhost:8765/teams/<team-id>/start

# Assign a task
curl -X POST http://localhost:8765/teams/<team-id>/tasks \
  -H "Content-Type: application/json" \
  -d '{"title": "Analyse data", "description": "Summarise Q4 revenue"}'

# Get events (Server-Sent Events)
curl http://localhost:8765/teams/<team-id>/events

Embedding in Another App​

Agent Teams can be mounted inside an existing FastAPI application:

from fastapi import FastAPI
from agent_teams.app import create_app
from agent_teams.manager import TeamManager

parent_app = FastAPI()
manager = TeamManager()

parent_app.mount("/api/agent-teams/v1", create_app(manager=manager))

Next Steps​