Agents

Creating Agents

Create your first agent in under two minutes — from the dashboard or by forking a template.

From the Dashboard

Go to Dashboard → Agents → New Agent

Give your agent a display name (e.g. "Research Assistant"). A URL-safe slug is generated automatically and becomes the agent's permanent ID (e.g. research-assistant).

Add steps

Every agent needs at least one step. A default "Planning & Reasoning" and "Main Task" step are added for you. Edit the prompts or add more steps as needed.

Enable tools & skills

Toggle built-in tools (Web Search, Calculator, etc.) and attach instructional skills from the Skills tab.

Test it

Use the Test tab in the agent editor. Type a message and click Run Test. You'll see each step execute in real time.

Agent ID

The agent ID is a URL-safe slug derived from the display name. It serves as both the unique identifier and the Firestore document ID.

Display NameMy Research Agent

Agent IDmy-research-agent

API Callmodel: "agent:my-research-agent"

If a slug already exists, a short random suffix is appended (e.g. my-research-agent-a3f2e1d4). The slug is immutable once created — like a GitHub repo name.

From a Template

Templates are pre-built agents you can fork into your account. They come with steps, skills, and settings already configured.

Chatbot Assistant

Productivity

Conversational Q&A with intent detection

Code Debugger

Coding

Analyze code, find bugs, suggest fixes

Content Summarizer

Content

Summarize articles with key points

Research Q&A

Productivity

Deep research with structured citations

Universal Translator

Communication

Context-aware translation between languages

Data Analyst

Data

Analyze datasets and generate insights

To fork a template: Dashboard → Agents → Templates, then click Fork on any template.

Calling Your Agent

Once your agent is active, call it via the OpenAI SDK pointed at the /api/agents endpoint.

javascript

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'mp_YOUR_API_KEY',
  baseURL: 'https://agentlify.co/api/agents',
});

const completion = await client.chat.completions.create({
  model: 'my-research-agent',
  messages: [{ role: 'user', content: 'What are the latest trends in AI?' }],
});

console.log(completion.choices[0].message.content);

Pre-launch Checklist

Agent has at least one enabled step with a prompt
Agent is set to Active (toggle in the editor header)
A router is linked (for model selection on LLM steps)
API key generated (Dashboard → Settings → API Keys)
Test run passes in the Dashboard Playground (Agent mode)

Copy the agent ID from the editor

In the agent editor header, click the copy icon next to the agent ID to get the slug for your API calls.

Next: Steps & Prompts

Learn how to write prompt templates and chain steps together.