Building Your First AI Agent
A Beginner’s Guide (No Coding Required)

Welcome to the World of AI Agents

An AI Agent is like your intelligent assistant — it can understand your requests, think about them, and perform actions. With the Arkus AI Agent Builder, you don’t need to write code. You build by connecting visual blocks, like putting together LEGO pieces.

You’ll learn:

• What an AI agent is
• How to design one step by step
• How to test and use it for real tasks

1. What Is an AI Agent?

Think of an agent as a person who works for you:

• You give it instructions (like: “Help me schedule a meeting”)
• It can use tools (like Google Calendar or email)
• It can read information (like files or guidelines)
• And it talks back to you in natural language!

Each agent you create has four main parts:

Part	What it does	Example
Chat Input	Where the user types a question	“Can you book a meeting with Dr. Andersson?”
Agent	The brain that thinks and decides what to do	Understands and plans the next action
Tool / File	Connects to external tools or documents	Google Calendar, or medical guidelines
Chat Output	Where you see the agent’s answer	“You’re free Tuesday at 10:00 or Thursday at 14:30.”

2. Understanding the Building Blocks

Each block (or node) is like one step in your agent’s thinking process. By connecting blocks, you define your agent’s workflow — how it receives messages, reasons, and responds.

There are two main agent types:

• Tool-based agent – connects to apps like Calendar, Email, CRM, etc.
• Knowledge-based agent – answers questions from uploaded files.

You can even combine both later!

3. Getting Ready

Before you start building:

1. Create your accounts:
   • Arkus AI Agent Builder → https://langflow.arkus.ai
   • OpenAI (for the brain) → https://platform.openai.com
   • Composio (for tools) → https://platform.composio.dev
     
     
2. Get your API keys (think of them as “keys” that unlock access for your agent).

4. Build Your First Agent — The Scheduling Assistant

This example teaches how to connect your agent to a real tool (Google Calendar).

Step 1: Create a New Flow

• Click “+ New Flow”
• Name it Scheduling Assistant
• You’ll see an empty workspace — that’s your agent’s world.

Step 2: Add the Building Blocks

Add these blocks:

• Agent
• Chat Input
• Chat Output
• Google Calendar

Connect them:

Chat Input → Agent → Chat Output
and Google Calendar → Agent (Tools dot)

This means: the user sends a message → the Agent thinks → uses the Calendar → sends back a reply.

Step 3: Configure the Agent

Click the Agent block, then:

• Paste your OpenAI API key
• In Instructions, describe what your agent does, for example:

You are a professional scheduling assistant.
Help users find suitable times, confirm details, and create meetings in Google Calendar.
Use a friendly and professional tone.

Step 4: Connect Google Calendar

Click the Google Calendar block, then:

• Paste your Composio API key
• Turn on Tool Mode (so the agent can take action)
• Approve the Google connection when asked
  

Step 5: Test It!

Click Playground → type:

“Can you schedule a team meeting next week?”
Watch your agent respond, ask questions, and book a slot.

Congratulations — you just built your first tool-connected agent!

5. Build a Knowledge Agent — The Guideline Assistant

This example shows how to make an agent that answers questions from documents.

Step 1: Create a New Flow

Name it Guideline Assistant.

Step 2: Add the Blocks

• Agent
• Chat Input
• Chat Output
• File (or PDF Loader)

Connect them:
Chat Input → Agent → Chat Output
and File → Agent (Knowledge dot)

Step 3: Upload Files

Click the File block, then upload a document (like a PDF guideline).

Step 4: Configure the Agent

Paste your OpenAI key and add instructions such as:

You are a guideline assistant.
Use uploaded documents to answer questions accurately.
If the information isn’t found, say so clearly.

Then click Playground and ask:

“What are the screening recommendations for patients under 40?”

Your agent will search the file and summarize the answer.

6. Tips for Success

• Start with one tool or one file, then expand.
• Always test with simple examples first.
• You can duplicate working agents to create new ones faster.
• Describe your agent’s instructions clearly — they shape its “personality.”
  

7. What You Can Build Next

Here are a few ideas:

• Clinical Assistant → reads medical guidelines and drafts patient summaries.
• Team Scheduler → manages meetings and reminders.
• Research Helper → summarizes research papers.
• Consent Form Builder → guides patients through digital consent steps.
  

Remember

You don’t need to know how to code. Just think about what your agent should do, what tools or files it needs, and how it should talk — then connect the right blocks.

Once your first agent works, you’re already thinking like an AI designer!

If you need a more detailed explanation, click here to read our Onboarding Manual.