Agents

Agents are the core building blocks of Deeployd. Each agent is an AI worker with a specific role, personality, and capabilities.

What is an Agent?

An agent is an AI assistant configured to:

• Have a specific role and personality
• Use certain tools and integrations
• Remember context through memory
• Work with other agents in teams

Think of agents as AI employees that you hire, train, and deploy.

Creating an Agent

Via Dashboard

1. Go to Dashboard → Agents → + Create Agent
2. Fill in the required fields
3. Click Create

Via API

import { DeepLoyd } from '@deeployd/sdk';

const client = new DeepLoyd({ apiKey: process.env.DEEPLOYD_API_KEY });

const agent = await client.agents.create({
  name: 'Support Agent',
  description: 'Handles customer support inquiries',
  systemPrompt: `You are a helpful customer support agent.
    Be friendly, professional, and solution-oriented.`,
  model: 'claude-3-5-sonnet-20241022',
  tools: ['memory', 'http_request']
});

console.log('Created agent:', agent.id);

Agent Configuration

Basic Settings

Field	Description	Required
name	Display name	Yes
description	What the agent does	No
systemPrompt	Personality and rules	Yes
model	AI model to use	No (default: Sonnet)

Model Options

Model	ID	Best For
Claude 3.5 Sonnet	claude-3-5-sonnet-20241022	General use
Claude 3 Opus	claude-3-opus-20240229	Complex reasoning
Claude 3 Haiku	claude-3-haiku-20240307	Fast, simple tasks

Advanced Settings

const agent = await client.agents.create({
  name: 'Advanced Agent',
  systemPrompt: '...',

  // Model settings
  model: 'claude-3-5-sonnet-20241022',
  temperature: 0.7,        // 0-1, higher = more creative
  maxTokens: 4096,         // Max response length

  // Capabilities
  tools: ['memory', 'calculator', 'http_request'],

  // Memory
  memory: {
    enabled: true,
    userMemory: { enabled: true, retention: '90d' },
    teamMemory: { enabled: true }
  },

  // Security
  systemPromptProtected: true,  // Prevent prompt injection

  // Team placement
  teamId: 'team-123'
});

Managing Agents

Get Agent

const agent = await client.agents.get('agent-123');

Update Agent

await client.agents.update({
  agentId: 'agent-123',
  name: 'Updated Name',
  systemPrompt: 'Updated prompt...'
});

List Agents

const agents = await client.agents.list({
  teamId: 'team-123',  // Optional: filter by team
  limit: 10,
  offset: 0
});

Delete Agent

await client.agents.delete('agent-123');

Chatting with Agents

Basic Chat

const response = await client.agents.chat({
  agentId: 'agent-123',
  message: 'Hello, how can you help me?'
});

console.log(response.content);

With Conversation Context

// Start a conversation
const response1 = await client.agents.chat({
  agentId: 'agent-123',
  message: 'My name is Sarah'
});

// Continue the conversation
const response2 = await client.agents.chat({
  agentId: 'agent-123',
  conversationId: response1.conversationId,
  message: 'What is my name?'
});
// Response: "Your name is Sarah"

With User Context

// Associate with a user for memory persistence
const response = await client.agents.chat({
  agentId: 'agent-123',
  userId: 'user-456',
  message: 'Remember that I prefer bullet points'
});

Streaming Responses

const stream = await client.agents.chat({
  agentId: 'agent-123',
  message: 'Tell me a story',
  stream: true
});

for await (const chunk of stream) {
  process.stdout.write(chunk.content);
}

Agent States

Agents can be in different states:

State	Description	Can Chat?
active	Normal operation	Yes
paused	Temporarily disabled	No
maintenance	Being updated	No
archived	Soft deleted	No

// Pause an agent
await client.agents.pause('agent-123');

// Resume an agent
await client.agents.resume('agent-123');

// Archive an agent
await client.agents.archive('agent-123');

Agent Relationships

Parent-Child (Escalation)

// Create child agent with parent
const childAgent = await client.agents.create({
  name: 'Tier 1 Support',
  systemPrompt: '...',
  parentAgentId: 'tier-2-agent-id',
  escalationRules: {
    conditions: ['user requests escalation', 'issue too complex'],
    message: 'Let me transfer you to a senior support agent.'
  }
});

Team Membership

// Add agent to team
await client.agents.addToTeam({
  agentId: 'agent-123',
  teamId: 'team-456'
});

// Move agent to different team
await client.agents.move({
  agentId: 'agent-123',
  toTeamId: 'team-789'
});

Agent Templates

Save and reuse agent configurations:

Save as Template

const template = await client.templates.createFromAgent({
  agentId: 'agent-123',
  name: 'IT Support Template',
  description: 'Template for IT support agents'
});

Create from Template

const agent = await client.agents.createFromTemplate({
  templateId: 'template-123',
  name: 'New IT Support Agent',
  customizations: {
    // Override template values
    systemPrompt: 'Custom additions...'
  }
});

Best Practices

Agent Design

1. Single Responsibility - Each agent should have one clear purpose
2. Clear Boundaries - Define what the agent can and cannot do
3. Descriptive Names - Use names that indicate the agent's role
4. Test Thoroughly - Test with realistic scenarios before deploying

System Prompts

✅ Good: Specific, actionable instructions

You are an IT Helpdesk agent for Acme Corp.
Your responsibilities:
- Answer IT questions
- Guide troubleshooting
- Create tickets when needed

Your boundaries:
- Don't access production systems
- Verify identity for password resets
- Escalate security concerns


❌ Bad: Vague, no boundaries

You are a helpful assistant.
Help users with their problems.

Error Handling

try {
  const response = await client.agents.chat({
    agentId: 'agent-123',
    message: 'Hello'
  });
} catch (error) {
  if (error.code === 'AGENT_PAUSED') {
    console.log('Agent is temporarily unavailable');
  } else if (error.code === 'RATE_LIMITED') {
    console.log('Too many requests, retry later');
  } else {
    throw error;
  }
}

Analytics

View agent performance:

const analytics = await client.agents.getAnalytics({
  agentId: 'agent-123',
  startDate: '2024-01-01',
  endDate: '2024-01-31'
});

console.log('Total conversations:', analytics.conversations);
console.log('Messages handled:', analytics.messages);
console.log('Avg response time:', analytics.avgResponseTime);
console.log('Token usage:', analytics.tokensUsed);

---

Next: Learn about System Prompts to master agent personality design.