Teams - PraisonAI

Agents can work together as a team to solve complex problems.

Quick Start

Simple Usage

import { Agent, Team } from 'praisonai';

const researcher = new Agent({
  name: 'Researcher',
  instructions: 'You research topics thoroughly'
});

const writer = new Agent({
  name: 'Writer',
  instructions: 'You write clear, engaging content'
});

const team = new Team({
  agents: [researcher, writer],
  process: 'sequential'
});

await team.start('Write an article about AI');
// Researcher gathers info → Writer creates article

With Configuration

const team = new Team({
  agents: [researcher, writer],
  process: 'hierarchical',
  manager: new Agent({ instructions: 'You coordinate the team' })
});

User Interaction Flow

Configuration Levels

// Level 1: Array - Simple team
const team = new Team({
  agents: [researcher, writer]
});

// Level 2: String - Named process
const team = new Team({
  agents: [researcher, writer],
  process: 'sequential'  // or 'parallel', 'hierarchical'
});

// Level 3: Dict - Full configuration
const team = new Team({
  agents: [researcher, writer],
  process: 'hierarchical',
  managerLlm: 'gpt-4o',       // model for the hierarchical manager
  memory: true,               // one shared store for the team
  planning: true,             // plan first, then run the steps
  execution: 'balanced',      // iteration/retry limits
  hooks: {                    // team lifecycle callbacks
    onTaskComplete: (task, output) => {},
  },
  verbose: true
});

runOn is refused on a team — passing it throws a TypeError at construction, matching Python. Put runOn on an individual Agent instead. See AgentTeam for the full list of team-level options.

Team Processes

Process	How It Works
`sequential`	Agents work one after another
`parallel`	Agents work at the same time
`hierarchical`	Manager coordinates agents

API Reference

TeamStructure

Team configuration

MultiAgentExecutionConfig

Multi-agent execution options

Best Practices

Give each agent one role

Specialized agents produce better results than generalists.

Use sequential for dependencies

When one agent needs another’s output, use sequential process.

Use hierarchical for complex tasks

A manager helps coordinate when tasks require judgment calls.

Python-Parity Aliases

To keep Python examples copy-pasteable, PraisonAIAgents and Agents are silent aliases of AgentTeam — same constructor, same behaviour.

import { Agent, PraisonAIAgents } from 'praisonai';

const researcher = new Agent({ name: 'Researcher', instructions: 'Research the topic.' });
const writer = new Agent({ name: 'Writer', instructions: 'Write the report.' });

// PraisonAIAgents === AgentTeam
const team = new PraisonAIAgents({ agents: [researcher, writer], process: 'sequential' });
await team.start('Write a report on renewable energy');

When both model and llm are set on a team, model wins (Python parity). Prefer one or the other to avoid confusion.

Handoffs

Transfer between agents

Workflows

Multi-step workflows

​Quick Start

​User Interaction Flow

​Configuration Levels

​Team Processes

​API Reference

TeamStructure

MultiAgentExecutionConfig

​Best Practices

​Python-Parity Aliases

​Related

Handoffs

Workflows

Quick Start

User Interaction Flow

Configuration Levels

Team Processes

API Reference

Best Practices

Python-Parity Aliases

Related