Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.praison.ai/llms.txt

Use this file to discover all available pages before exploring further.

AgentTeam is the multi-agent coordinator that manages and delegates work to multiple Agent instances, supporting sequential, parallel, and hierarchical execution patterns.

​
Quick Start

1

Basic Team

Create a simple team with multiple agents:
from praisonaiagents import Agent, AgentTeam, Task

# Create agents
researcher = Agent(
    name="Researcher",
    role="Research Specialist",
    instructions="Research and gather information on topics"
)

writer = Agent(
    name="Writer", 
    role="Content Writer",
    instructions="Write clear, engaging content"
)

# Create tasks
task1 = Task(description="Research AI trends", agent=researcher)
task2 = Task(description="Write article", agent=writer)

# Create team
team = AgentTeam(
    agents=[researcher, writer],
    tasks=[task1, task2]
)

result = team.start()
print(result)
2

Parallel Execution

Execute tasks in parallel for faster processing:
from praisonaiagents import Agent, AgentTeam, Task

agents = [
    Agent(name="Analyst1", instructions="Analyze market data"),
    Agent(name="Analyst2", instructions="Analyze competitor data"),
    Agent(name="Analyst3", instructions="Analyze customer data"),
]

tasks = [
    Task(description="Market analysis", agent=agents[0]),
    Task(description="Competitor analysis", agent=agents[1]),
    Task(description="Customer analysis", agent=agents[2]),
]

team = AgentTeam(
    agents=agents,
    tasks=tasks,
    process="parallel"  # Execute tasks in parallel
)

result = team.start()
3

Hierarchical Process

Use a manager agent to coordinate work:
from praisonaiagents import Agent, AgentTeam, Task

team = AgentTeam(
    agents=[researcher, writer, editor],
    tasks=[research_task, write_task, edit_task],
    process="hierarchical",  # Manager coordinates agents
    manager_llm="gpt-4o"     # Manager model
)

result = team.start()
4

With Memory & Planning

Enable shared memory and planning for the team:
from praisonaiagents import Agent, AgentTeam

team = AgentTeam(
    agents=[agent1, agent2, agent3],
    memory=True,      # Shared memory across agents
    planning=True,    # Enable collaborative planning
    output="verbose"  # Detailed output
)

result = team.start("Analyze and report on Q4 sales data")

​
Process Types

ProcessDescriptionBest For
sequentialTasks execute one after anotherDependent tasks, pipelines
parallelTasks execute simultaneouslyIndependent analysis
hierarchicalManager coordinates agentsComplex delegation

​
Configuration Options

ParameterTypeDefaultDescription
agentsList[Agent]RequiredList of Agent instances
tasksList[Task]Auto-generatedTasks to execute
processstr"sequential"Execution pattern
manager_llmstr"gpt-4o-mini"Manager model (hierarchical)
memorybool/ConfigFalseShared memory system
planningbool/ConfigFalseCollaborative planning
contextbool/ConfigFalseContext management
outputstr/ConfigNoneOutput configuration
executionstr/ConfigNoneExecution limits
hooksConfigNoneLifecycle callbacks

​
State Management

AgentTeam provides shared state across all agents: 
team = AgentTeam(agents=[agent1, agent2])

# Set shared state
team.set_state("analysis_complete", True)
team.set_state("total_records", 1500)

# Get state from any agent
value = team.get_state("analysis_complete")

# Increment counters
team.increment_state("processed_count", 1)

# Append to lists
team.append_to_state("findings", "New insight discovered")

​
Best Practices

For sequential tasks, AgentTeam automatically passes context from previous tasks:
task1 = Task(description="Research topic", agent=researcher)
task2 = Task(description="Write based on research", agent=writer)
# task2 automatically receives task1's output as context
Create agents with distinct roles for better task delegation:
researcher = Agent(role="Research Specialist", tools=[search_tool])
analyst = Agent(role="Data Analyst", tools=[analysis_tool])
writer = Agent(role="Content Writer")
Enable memory for teams that need to share information:
team = AgentTeam(
    agents=[...],
    memory=True  # All agents share memory
)

​
Key Methods

MethodDescription
start()Execute the team workflow
run()Alias for start()
astart()Async execution
set_state()Set shared state value
get_state()Get shared state value
add_task()Add task dynamically
get_task_status()Check task status

Agent

Learn about individual AI agents

AgentFlow

Deterministic workflow pipelines

Tasks

Task configuration and management

Process

Execution process patterns