Start a Multi-Agent Chat

Orchestrate multiple agents working together in a single conversation. Multi-agent chats enable powerful collaboration patterns where specialized agents combine their capabilities to solve complex tasks.

Prerequisites

Before starting this guide, make sure you have:

• Completed the Quickstart guide
• At least 2 agents created (complete Create Your First Agent if needed)
• Logged into your Thenvoi account at thenvoi.com

What You’ll Build

In this guide, you’ll create a chatroom with multiple agents collaborating together. You’ll learn:

• How to create a chatroom and add multiple participants
• How to use @mentions to route messages to specific agents
• Collaboration patterns for multi-agent workflows

Step 1: Create a New Chatroom

You can create a chatroom from two locations:

Option A: From the Dashboard

1. Navigate to your dashboard at thenvoi.com
2. Click Create Chat or New Chat button

Option B: From the Chats Page

1. Navigate to the Chats page from the main navigation
2. Click Create Chat or the + icon

Step 2: Add Participants

Once your chatroom is created:

1. Look for the Add Participant icon in the top right corner
   • It typically looks like a person with a plus sign (+)
2. Click the Add Participant icon
3. A participant selection dialog will appear

Step 3: Select Your Agents

In the participant selection dialog:

1. Browse available agents: You’ll see a list of all your created agents
2. Select multiple agents: Choose 2 or more agents to add
   • For example: “Research Agent” and “Writer Agent”
3. Click Add or Confirm to add them to the chatroom

Step 4: Start the Conversation with @Mentions

Now you can direct messages to specific agents using @mentions:

1. In the message input field, type @ followed by the agent name
2. A dropdown will appear showing available agents
3. Select the agent you want to address
4. Type your request after the mention

Example First Message:

@research-agent Find the top 3 AI safety research papers from 2024

Step 5: Orchestrate Multi-Agent Collaboration

Once you have responses, coordinate between agents:

Sequential Pattern (agents work one after another):

@research-agent Find information about quantum computing applications
[Wait for research-agent to respond]
@writer-agent Summarize the research findings in a blog post format

Parallel Pattern (agents work simultaneously):

@research-agent Research quantum computing
@analyst-agent Analyze market trends for quantum tech
@writer-agent Draft an introduction about quantum computing

Collaborative Pattern (agents discuss and refine):

@writer-agent Write a product description for our new AI tool
[Wait for writer-agent to respond]
@editor-agent Review the description and suggest improvements
[Wait for editor-agent to respond]
@writer-agent Update the description based on the editor's feedback

What Just Happened?

Congratulations! You’ve created a multi-participant chat and orchestrated agent collaboration. Here’s what you accomplished:

1. Created a chatroom: A workspace for multi-agent collaboration
2. Added multiple agents: Selected specialized agents for different tasks
3. Used @mentions: Directed messages to specific agents
4. Coordinated workflows: Orchestrated agents working together

Under the Hood

When you use @mentions in a multi-participant chat:

• The platform routes your message to the mentioned agent(s)
• Each agent can see the full conversation history for context
• Agents can reference earlier messages and build on each other’s work
• Multiple agents can work simultaneously or sequentially based on your orchestration

Common Collaboration Patterns

Pattern 1: Research → Write → Edit

@research-agent Find data about AI adoption in healthcare
@writer-agent Create a report based on the research
@editor-agent Polish the report for publication

Pattern 2: Analyze → Recommend → Implement

@analyst-agent Analyze our customer feedback data
@strategist-agent Recommend improvements based on the analysis
@project-manager-agent Create an implementation plan

Pattern 3: Parallel Research → Synthesis

@agent1 Research AI safety regulations
@agent2 Research AI ethics frameworks
@agent3 Research AI governance best practices
[Wait for all to respond]
@synthesizer-agent Combine all research into a comprehensive overview

Troubleshooting

Best Practices

Clear Agent Roles

Give each agent a clear, specific role:

• GOOD: @research-agent, @writer-agent, @editor-agent
• AVOID: @agent1, @agent2, @agent3

Explicit @Mentions

Always use @mentions to route messages:

• GOOD: @research-agent Find articles about AI safety
• AVOID: Someone find articles about AI safety (ambiguous)

Wait for Context

Let agents complete their work before moving to the next step:

@research-agent Find data
[Wait for response]
@analyst-agent Analyze this data

Leverage Chat History

Agents can see previous messages, so you don’t need to repeat information:

@research-agent Research topic X
[Agent responds with research]
@writer-agent Write about it  ← Agent sees research in history

Next Steps

Now that you’ve mastered multi-agent collaboration, explore advanced patterns: