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.
Specialized agents, Taskade EVE, and people coordinate around one shared project.
graph TD
HUMAN["Human team members<br/>direction and decisions"] --> WS["Shared project / workspace"]
WS --> EVE["Taskade EVE<br/>orchestrates and invokes agents"]
EVE --> RES["Research Agent"]
EVE --> ANL["Analyst Agent"]
EVE --> WRT["Writer Agent"]
EVE --> COORD["Coordinator Agent"]
RES --> WS
ANL --> WS
WRT --> WS
COORD --> WS
Workspace DNA impact: Multi-agent teams amplify the Intelligence pillar. Each agent contributes +10 points to your Intelligence Score, and teams create compounding intelligence through specialization.
Plan note: Multi-Agent Teams are available on Business+ plans.
| 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 |
| 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 |
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 |
| 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 |
| 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 |
| Step | Action |
|---|
| 1 | Select multiple tasks or projects |
| 2 | Use bulk actions to assign agents |
| 3 | Agents are assigned to all selected items |
| 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 |
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" |
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.
| 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) |
| 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 Taskade Genesis app | App users | Add agent chat widget to your Genesis app |
| 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 |