Create Your First Agent

By now you have a sense of what Zia Agents is and what type of agent fits your goal. This guide walks you through actually building one. We'll cover all four configurations so you can follow whichever path matches your use case.

The entire process happens inside Agent Studio, and the steps are largely the same regardless of which type you're creating. The only difference is whether you add a Knowledge Base, Tools, both, or neither along the way.

Before you begin

Every agent starts with the same foundation: a set of basic information that tells the platform who the agent is and how it should behave. You'll fill this in regardless of which type you're building, so it's worth understanding what each field does before you start.

When you open Agent Studio and choose Create From Scratch (or use Create with Zia AI if you'd prefer an assisted setup), you'll be asked to provide the following:

Agent name and description: What the agent is called and a short summary of its purpose. These are visible to anyone browsing your agent list, so keep them clear.
Vendor configuration and model: Which LLM provider and model the agent should use. Your choice here depends on your organization's preferences around capability, cost, and data handling.
Role: A short statement of what the agent's job is. Think of it as the agent's job title in one or two lines.
Instructions: This is where you define how the agent should behave, what tone to use, what topics to cover, what to avoid, and any specific logic it should follow. The more specific you are here, the better the agent performs.

Once you've filled this in and clicked Continue, you'll move through the knowledge base, tools, and guardrails steps. What you do at each of those steps is what determines your agent type.

Type 1: Agent without tools or knowledge base

The simplest path. Your agent will rely on the language model's general knowledge and follow the instructions you've provided.

Complete the basic information as described above and click Continue.
On the Knowledge Base step, skip the selection and click Continue.
On the Tools step, skip the selection and click Continue.
On Additional Settings, configure guardrails if needed and click Create.
Your agent is ready to test.

Type 2: Agent with knowledge base only

This gives your agent access to your own documents so its responses are grounded in your business content rather than general knowledge.

Complete the basic information and click Continue.
On the Knowledge Base step, select the documents relevant to your use case and click Continue.
These can be product docs, FAQs, internal guides, or any reference material you've uploaded to the Knowledge Base section of the platform.
On the Tools step, skip the selection and click Continue.
On Additional Settings, configure guardrails if needed and click Create.

Type 3: Agent with tools only

This configuration lets the agent perform actions in your systems through API connections. It uses the model's reasoning to decide when and how to use the tools you've provided.

Complete the basic information and click Continue.
On the Knowledge Base step, skip the selection and click Continue.
On the Tools step, click + Add Tools.
You'll see two categories: System Tools (pre-built integrations with Zoho products) and Custom Tools (API connections you've configured yourself). Select what the agent needs, click Add, then Continue.
On Additional Settings, configure guardrails if needed and click Create.

Type 4: Agent with tools and knowledge base

The full setup. Your agent gets both business context from your documents and the ability to act through connected APIs.

Complete the basic information and click Continue.
On the Knowledge Base step, select the relevant documents and click Continue.
On the Tools step, click + Add Tools, select the tools you need, click Add, then Continue.
On Additional Settings, configure guardrails if needed and click Create.

What to do after creation

Once your agent is created, you'll land on the agent details page. For a breakdown of everything on this page, see Agents.

Before you deploy it anywhere, take it for a test run.

Click Test Agent in the top right corner to open the test window. This lets you interact with the agent in real time and see exactly how it responds based on the configuration you've set up.
If something feels off, whether it's the tone, the accuracy of responses, or how it uses tools, you can go back and adjust the instructions, swap out knowledge base documents, or reconfigure the tools.
Testing is covered in detail in the Testing Agents article, but the short version is: don't skip it. Even a well-configured agent benefits from a few rounds of testing before it goes live.