Where Each Option Belongs

AgentTeam (also exported as PraisonAIAgents) applies autonomy at the team level; the feature options belong on each Agent.

execution follows the same split: pass MultiAgentExecutionConfig at the team level and the single-agent ExecutionConfig on each Agent. Passing the single-agent ExecutionConfig to the team now emits a UserWarning listing the fields it cannot map. See ExecutionConfig vs MultiAgentExecutionConfig.

Quick Start

Put feature options on each Agent

Give every agent its own web, knowledge, or other feature options.

from praisonaiagents import Agent, AgentTeam

researcher = Agent(
    name="Researcher",
    instructions="Research the topic.",
    web=True,                    # Feature options live on the Agent
    knowledge=["notes.txt"],
)
writer = Agent(
    name="Writer",
    instructions="Write a summary.",
    reflection=True,
)

team = AgentTeam(agents=[researcher, writer])
team.start("Summarise the latest AI safety research")

Put autonomy on the team

autonomy is the one option the team applies for you.

from praisonaiagents import Agent, AgentTeam

team = AgentTeam(
    agents=[Agent(name="Worker", instructions="Do the task.")],
    autonomy=True,               # Applied at the team level
)
team.start("Plan and execute the task")

How It Works

AgentTeam accepts six feature options for API symmetry with Agent, but does not apply them at the team level. Passing them logs a warning and ignores the value.

Option	Belongs on	Passing it to the team does…
`autonomy`	`AgentTeam`	Applies at the team level
`guardrails`	Each `Agent`	Warns and is ignored at the team level
`web`	Each `Agent`	Warns and is ignored at the team level
`reflection`	Each `Agent`	Warns and is ignored at the team level
`caching`	Each `Agent`	Warns and is ignored at the team level
`learn`	Each `Agent`	Warns and is ignored at the team level
`knowledge`	Each `Agent`	Warns and is ignored at the team level

Passing guardrails, web, reflection, caching, learn, or knowledge to AgentTeam / PraisonAIAgents triggers:

AgentTeam received ['web'] but does not yet apply them at the team level;
pass these to individual Agent(...) instances instead.

Common Patterns

Move the option down from the team onto each agent.

from praisonaiagents import Agent, AgentTeam

team = AgentTeam(
    agents=[
        Agent(name="Researcher", instructions="Research."),
        Agent(name="Writer", instructions="Write."),
    ],
    web=True,          # Warns and is ignored
)

from praisonaiagents import Agent, AgentTeam

team = AgentTeam(
    agents=[
        Agent(name="Researcher", instructions="Research.", web=True),
        Agent(name="Writer", instructions="Write.", web=True),
    ],
)

Best Practices

Set feature options per role

Each agent has a different job. Give the researcher web=True and the writer reflection=True rather than forcing one option across the whole team.

Reserve the team for autonomy

autonomy is the only feature option the team propagates. Keep team-level configuration to autonomy, process, and manager_llm.

Watch the logs

A warning on construction means an option landed in the wrong place. Move it onto the Agent(...) instances it was meant for. Passing the single-agent ExecutionConfig to the team warns and adapts what it can — use MultiAgentExecutionConfig instead. See ExecutionConfig vs MultiAgentExecutionConfig.

Knowledge Quick Start

Give each agent its own documents

Multi-Agent Pipelines

Coordinate multiple agents

Getting Started

Install

Core Concepts

Guides

Features

Models

Databases

Observability

Memory

Telemetry & Analytics

Knowledge

RAG

Persistence

Tools

Other Features

Developers

Configuration

Best Practices

Getting Started (No Code)

Where Each Option Belongs

Quick Start

How It Works

Common Patterns

Best Practices

Knowledge Quick Start

Multi-Agent Pipelines

​Quick Start

​How It Works

​Common Patterns

​Best Practices

​Related

Knowledge Quick Start

Multi-Agent Pipelines

Quick Start

How It Works

Common Patterns

Best Practices

Related