# Multi-Agent Collaboration

***

## Table of Contents

* [Overview](#overview)
* [Set Up Your AI Agents](#set-up-your-ai-agents)
* [Team Up with AI Agents](#team-up-with-ai-agents)
* [Use the Agent Sidebar](#use-the-agent-sidebar)
* [Iterate on Output](#iterate-on-output)
* [Chat with Agents & Your Team](#chat-with-agents--your-team)
* [Copy / Duplicate Multi-Agents](#copy--duplicate-multi-agents)
* [Share Agents](#share-agents)
* [Multi-Agent Architecture Patterns](#multi-agent-architecture-patterns)
* [What's Next](#whats-next)

***

## Overview

Multi-Agents lets you **assign multiple AI agents to work together** on projects, tasks, and conversations. Instead of relying on a single generalist AI, you create teams of specialized agents that collaborate — each bringing domain expertise, unique knowledge, and specific tools.

> **Workspace DNA impact:** Multi-agent teams amplify the Intelligence pillar. Each agent contributes +10 points to your [Intelligence Score](https://docs.taskade.com/docs/taskade-genesis/genesis/how-genesis-works#workspace-intelligence-score), and teams create compounding intelligence through specialization.
>
> **Plan note:** Multi-Agent Teams are available on Business+ plans.

***

## Set Up Your AI Agents

### Method 1: Build from Scratch

| Step | Action                                           |
| ---- | ------------------------------------------------ |
| 1    | Open the **Agents** tab in your workspace        |
| 2    | Click **Create Agent**                           |
| 3    | Configure: name, persona, tone, knowledge, tools |
| 4    | Repeat for each specialized agent in your team   |

### Method 2: Use Templates

| Step | Action                                          |
| ---- | ----------------------------------------------- |
| 1    | Browse the agent template library               |
| 2    | Select pre-built agent templates for each role  |
| 3    | Customize each agent for your specific use case |

### Method 3: Auto-Generate a Team with AI

The fastest way to create a multi-agent team:

| Step | Action                                                                                                                                                   |
| ---- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
| 1    | Open the **Agents** tab                                                                                                                                  |
| 2    | Click **Create Agent Team**                                                                                                                              |
| 3    | Describe your team in plain English (e.g., "I need a marketing team with a content writer, SEO specialist, social media manager, and analytics analyst") |
| 4    | AI generates the complete team with specialized agents, knowledge sources, and tool configurations                                                       |
| 5    | Review and customize each agent as needed                                                                                                                |

> **Pro tip:** Auto-generated teams include appropriate personas, knowledge connections, and tool assignments based on your description. You can always fine-tune each agent individually afterward.

**Example team setup:**

| Agent                 | Specialty                     | Knowledge                         | Tools                                   |
| --------------------- | ----------------------------- | --------------------------------- | --------------------------------------- |
| **Research Agent**    | Data gathering, web search    | Industry reports, competitor URLs | Search Web, Scrape Webpage              |
| **Analyst Agent**     | Data analysis, insights       | Historical data projects, KPIs    | Google Sheets, Custom Fields            |
| **Writer Agent**      | Content creation, copywriting | Brand guidelines, tone docs       | Google Docs, WordPress                  |
| **Coordinator Agent** | Task management, routing      | Team structure, processes         | Slack, Google Calendar, Taskade Actions |

***

## Team Up with AI Agents

### Assign a Single Agent

| Step | Action                                                   |
| ---- | -------------------------------------------------------- |
| 1    | Open a project or task                                   |
| 2    | Click the **Agent** icon                                 |
| 3    | Select an agent from your workspace                      |
| 4    | The agent now has context for that specific project/task |

### Assign Multiple Agents

| Step | Action                                          |
| ---- | ----------------------------------------------- |
| 1    | Open a project                                  |
| 2    | Click the **Agent** icon                        |
| 3    | Select multiple agents for the project          |
| 4    | Each agent applies its specialty to the project |

### Bulk-Assign Agents

| Step | Action                                    |
| ---- | ----------------------------------------- |
| 1    | Select multiple tasks or projects         |
| 2    | Use bulk actions to assign agents         |
| 3    | Agents are assigned to all selected items |

***

## Use the Agent Sidebar

| Feature               | Detail                                               |
| --------------------- | ---------------------------------------------------- |
| **Access**            | Click the Agent icon in the sidebar                  |
| **Switch agents**     | Toggle between assigned agents mid-conversation      |
| **Context awareness** | Agent sees the project/task you're currently viewing |
| **Tool access**       | All enabled tools are available in the sidebar       |
| **History**           | Conversation history preserved per agent             |

***

## Iterate on Output

Multi-agent output can be refined iteratively:

| Technique                 | Example                                                              |
| ------------------------- | -------------------------------------------------------------------- |
| **Sequential refinement** | Research Agent drafts → Writer Agent refines → Reviewer Agent checks |
| **Parallel generation**   | Multiple agents tackle different sections simultaneously             |
| **Feedback loop**         | Ask one agent to critique another agent's output                     |
| **Role switching**        | "Now have the Analyst Agent review what the Writer Agent produced"   |

***

## Chat with Agents & Your Team

Multi-agent chats support both AI agents and human team members in the same conversation:

| Participant            | Role                                           |
| ---------------------- | ---------------------------------------------- |
| **AI Agents**          | Respond based on knowledge and tools           |
| **Human team members** | Provide direction, make decisions, add context |
| **EVE**                | Orchestrates and can invoke other agents       |

> **Tip:** @mention specific agents in chat to direct questions or tasks to them.

***

## Copy / Duplicate Multi-Agents

| Action                      | How                                                 | Result                                                     |
| --------------------------- | --------------------------------------------------- | ---------------------------------------------------------- |
| **Copy an agent**           | Agent settings → Duplicate                          | Creates an identical copy you can customize                |
| **Copy a team**             | Duplicate the project with agent assignments        | All agents and their assignments are copied                |
| **Share across workspaces** | Publish agent publicly → Clone in another workspace | Agent config is replicated (knowledge may need re-linking) |

***

## Share Agents

| Sharing Method              | Scope            | How                                                      |
| --------------------------- | ---------------- | -------------------------------------------------------- |
| **Within workspace**        | Team members     | Agents are available to all workspace members by default |
| **Cross-workspace**         | Other workspaces | Publish publicly → Clone                                 |
| **Public**                  | Anyone           | Publish with public toggle                               |
| **Embedded in Genesis app** | App users        | Add agent chat widget to your Genesis app                |

***

## Multi-Agent Architecture Patterns

| Pattern                 | Description                                         | Best For                          | Example                                                                |
| ----------------------- | --------------------------------------------------- | --------------------------------- | ---------------------------------------------------------------------- |
| **Sequential Pipeline** | Agent A → Agent B → Agent C (output flows linearly) | Content creation, data processing | Research → Write → Review                                              |
| **Hub and Spoke**       | Central coordinator routes to specialists           | Customer support, complex queries | Coordinator → routes to Sales/Support/Billing                          |
| **Parallel Fan-Out**    | Multiple agents work simultaneously                 | Report generation, data analysis  | Finance + Marketing + Operations agents each analyze their domain      |
| **Debate / Review**     | Agents critique each other's work                   | Quality control, decision-making  | Proposal Agent creates → Critic Agent reviews → Proposal Agent revises |
| **Escalation Chain**    | Start with generalist, escalate to specialist       | Support tiers, triage             | Tier 1 Agent → Tier 2 Specialist → Human escalation                    |

***

## What's Next

| Guide                                                                                               | What You'll Learn                     |
| --------------------------------------------------------------------------------------------------- | ------------------------------------- |
| [Custom AI Agents](https://docs.taskade.com/docs/ai-features/ai-features/ai-agents-getting-started) | Build individual agents               |
| [Agent Knowledge & Memory](https://docs.taskade.com/docs/ai-features/ai-features/agent-knowledge)   | Train agents with domain knowledge    |
| [Tools for AI Agents](https://docs.taskade.com/docs/ai-features/ai-features/agent-tools)            | Add capabilities to each agent        |
| [Automations](https://docs.taskade.com/docs/automations/automation)                                 | Connect agents to automated workflows |