defi-agents

Agent Teams Guide

How to use multiple specialized agents collaboratively in AI Agents Library.

What Are Agent Teams?

Agent Teams bring together multiple AI agents, each with their own expertise, to collaborate on conversations. Instead of one agent’s perspective, you get comprehensive answers from a team of specialists.

Key Benefits:

How Agent Teams Work

The Flow

  1. You ask a question or present a problem
  2. Host agent determines which team members should respond
  3. Agents discuss and build on each other’s insights
  4. Host coordinates the conversation flow
  5. You receive comprehensive, multi-perspective answer

The Host Agent

Every team has a host who:

You can:

Creating an Agent Team

Quick Start

  1. Click “Create Agent Team” in AI Agents Library
  2. Choose method:
    • Use preset template (recommended for beginners)
    • Build custom team from your agents

From Template

Available Templates:

DeFi Strategy Team

Content Creation Team

Research Team

Problem Solving Team

Select template → Customize members if needed → Start conversation

Build Custom Team

  1. Select 3-5 agents from your workspace
  2. Assign roles (optional but recommended)
  3. Configure host behavior
  4. Set team name
  5. Save and start

Using Agent Teams

@ Mentions

Target specific team members:

@Portfolio-Analyzer what's my current allocation?

Only that agent responds directly.

Private Messages

Click agent’s avatar to send direct message:

Interrupt and Resume

Interrupt: Click “Pause” or interrupt naturally

Resume: Click “Continue” or prompt host

Team Speed

Adjust response pacing:

Example Scenarios

DeFi Portfolio Analysis

Team:

User: “Review my positions and suggest improvements”

Flow:

  1. Portfolio Analyzer examines holdings
  2. Risk Assessor evaluates exposure
  3. Yield Optimizer identifies opportunities
  4. Rebalancing Advisor suggests changes
  5. Host synthesizes recommendations

Result: Comprehensive analysis from multiple angles

Content Strategy

Team:

User: “Plan a content series about DeFi for beginners”

Flow:

  1. Researcher identifies trending topics
  2. Writer proposes article structures
  3. SEO suggests keywords and optimization
  4. Distribution recommends channels
  5. Host creates unified strategy

Result: Complete content plan ready to execute

Technical Problem

Team:

User: “Review this contract for issues”

Flow:

  1. Auditor scans for vulnerabilities
  2. Gas Optimizer finds efficiency improvements
  3. Security Expert assesses risks
  4. Best Practices suggests improvements
  5. Host prioritizes findings

Result: Thorough code review from multiple dimensions

Best Practices

Team Size

Optimal: 3-5 agents

Too few (1-2): Limited perspectives Too many (6+): Coordination overhead, slower responses

Agent Selection

Mix expertise types: ✅ Analyst + Strategist + Risk Assessor + Implementer ❌ 4 variations of the same role

Complementary skills:

Host Customization

Default host works for most cases

Customize when:

Advanced Options → Custom Moderator:

You are a DeFi-focused team coordinator.

Prioritize:
- Risk assessment first
- Security considerations always
- Gas costs in every recommendation

Guide discussion toward:
- Actionable recommendations
- Clear next steps
- Risk-adjusted thinking

Conversation Management

Start broad, get specific:

✅ "Analyze our DeFi strategy"
   → Let team discuss
   → Then: "@Risk-Assessor what's our worst-case scenario?"

❌ "Tell me exactly what to do"
   → Too directive
   → Doesn't leverage team dynamics

Use private messages for:

Use @ mentions for:

Advanced Features

Team Templates

Save your custom teams:

  1. Create and configure team
  2. Click “Save as Template”
  3. Name and describe
  4. Reuse anytime

Share templates:

Sequential vs Parallel

Sequential (Default):

Parallel (Advanced):

Role Assignment

Give agents specific responsibilities:

Portfolio Analyzer: "Data collector"
Risk Assessor: "Devil's advocate"
Yield Optimizer: "Opportunity spotter"
Host: "Decision maker"

Roles guide behavior and coordination.

Common Use Cases

1. Investment Analysis

Team: Fundamental Analyst + Technical Analyst + Risk Manager + Market Strategist

Use: Comprehensive investment evaluation

2. Content Production

Team: Researcher + Writer + Editor + SEO Specialist

Use: End-to-end content creation

3. Project Planning

Team: Strategist + Technical Lead + Resource Manager + Risk Assessor

Use: Complete project roadmap

4. Code Review

Team: Security Auditor + Performance Optimizer + Best Practices Expert + Documentation Reviewer

Use: Multi-dimensional code assessment

5. Decision Making

Team: Pros/Cons Analyst + Devil’s Advocate + Opportunity Spotter + Risk Evaluator

Use: Balanced decision analysis

Troubleshooting

Team responses too slow

Solutions:

Agents talk past each other

Solutions:

Too much agreement (no debate)

Solutions:

Host too dominant

Solutions:

Missing key perspective

Solutions:

Creating Agent Teams for Others

Team Design Principles

  1. Clear division of labor
  1. Appropriate size
  1. Balanced perspectives
  1. Host instructions

Example: Research Team

{
  "agents": [
    "literature-reviewer",
    "methodology-expert",
    "data-analyst",
    "citation-checker",
    "synthesizer"
  ],
  "host": {
    "instructions": "Coordinate academic research process. Ensure methodology rigor, proper citations, and clear synthesis. Facilitate from research question → literature review → methodology → analysis → conclusions."
  },
  "name": "Academic Research Team"
}

Tips for Power Users

Reusable Templates

Create templates for recurring workflows:

Hybrid Approach

Combine team discussion with individual agent queries:

  1. Get team perspective
  2. @ mention specific agent for depth
  3. Use private messages for follow-ups
  4. Return to team for synthesis

Iterative Refinement

First run: See how team interacts Second run: Adjust roles and host Third run: Optimize team composition Fourth+ run: Fine-tuned workflow

Documentation

Keep notes on what works:

FAQ

Q: Can I use the same agent twice in a team? A: Technically yes, but not recommended. Use different specialized agents instead.

Q: Do agents see each other’s full conversation history? A: Yes, that’s how they build on each other’s insights.

Q: Can I remove an agent mid-conversation? A: Not currently. Plan team composition before starting.

Q: Does team size affect cost? A: Yes—more agents = more tokens = higher cost. Keep teams lean.

Q: Can teams work on long documents? A: Yes, but monitor context limits. Teams consume more tokens than solo agents.

Resources

Agent Teams transform AI from solo performer to collaborative ensemble. Start simple, experiment, refine.

This site is open source. Improve this page.