Get Started

Build and deploy your first AI agent without installing anything. All you need is a browser and an account.

Prerequisites

• Latest web browser (Google Chrome, Mozilla Firefox, Apple Safari, or Microsoft Edge).
• An Agent Platform 2.0 account - sign up free at ablplatform.com.

Setup Guide

Set up your first multi-agent system in four phases:

Phase 1: Access Studio

Sign up and log in

Go to ablplatform.com and create your account. After verifying your email, you land in Studio — the browser-based IDE where you build, test, and manage your agents.

Create your first project

Click New Project from the Studio dashboard. Give it a name like “My First Agents” and select a workspace. Projects organize your agents, supervisors, tools, and knowledge sources in one place.

Phase 2: Build your First Agent

Inside your project, click New Agent. Open the ABL editor and paste this definition:

AGENT: Support_Assistant

EXECUTION:
  model: claude-sonnet-4-5-20250929

GOAL: |
  Help customers with product questions. Be concise
  and friendly. If you do not know the answer, say so.

PERSONA: |
  Helpful product support assistant. Answers questions
  clearly and concisely.

LIMITATIONS:
  - "Cannot process payments or refunds"
  - "Cannot access customer account information"

TOOLS:
  search_knowledge(query: string) -> {results: object[], totalCount: number}
    description: "Search the product knowledge base"

INSTRUCTIONS: |
  1. Understand the customer's question
  2. Search the knowledge base for relevant information
  3. Provide a clear, sourced answer
  4. If unsure, offer to connect with a human agent

This definition creates an agent that:

• Uses an LLM to understand customer questions.
• Searches a knowledge base for answers.
• Responds with sourced information.
• Has clear boundaries on what it can and cannot do.

Click Save to validate your definition. Studio parses ABL in real time and flags syntax issues inline.

Phase 3: Test your Agent

Open the Test panel on the right side of Studio and send a message:

What is your return policy?

Your agent processes the message, searches for relevant knowledge, and responds. The trace viewer below the chat shows the full execution — LLM calls, tool invocations, and reasoning steps.

Send a few more messages to see the agent handle different questions, maintain context across turns, and respect its defined limitations.

Phase 4: Add a Supervisor

Create a new Supervisor in your project and paste this definition:

SUPERVISOR: Product_Supervisor

EXECUTION:
  model: claude-sonnet-4-5-20250929

GOAL: |
  Route customer queries to the right specialist agent.

HANDOFF:
  - TO: Support_Assistant
    WHEN: user asks about products, features, or general help
    PASS: query

  - TO: Billing_Agent
    WHEN: user asks about invoices, payments, or subscriptions
    PASS: query

The supervisor evaluates each incoming message and routes it to the right agent, passing conversation context along. Test it the same way — open the Test panel and send messages that should route to different agents.

What Have You Built

In a few minutes, you created:

• An agent that understands natural language, retrieves knowledge, and enforces boundaries.
• A supervisor that routes messages to the right specialist.
• Observable traces for every execution step, visible right in Studio.

Next Step

• Your first agent tutorial — Build a complete agent with tools, knowledge, and guardrails.

Check out the Template Gallery in Studio for ready-made agent definitions across industries — airlines, retail, banking, telecom, travel, and more.

---