Skip to main content

Getting Started

How to Create an AI Agent in KriyaOS

This guide helps you create your first agent in KriyaOS.

It is written for first-time users and focuses on what you will actually see and decide in the product.

Quick Recommendation

If you want the fastest guided setup, start with Business Agent. If you want more control before anything goes live, choose Custom Agent.

What This Page Helps You Do

By the end of this page, you should understand:

  • the two ways to create an agent
  • which path is better for you
  • what the main setup fields mean
  • what happens after creation
  • what to watch out for as a new user

Before You Begin

Before you create an agent, it helps to know two things:

  1. Which workspace you want to create it in
  2. Whether you want a guided setup or a more hands-on setup

If you plan to use uploaded knowledge, keep in mind that files may need a little time to finish processing before the agent can use them fully.

Choose The Right KriyaOS Agent Creation Path

KriyaOS currently gives you two creation paths:

  • Business Agent A guided onboarding flow that helps you move faster. It creates and publishes the agent for you as part of the flow.

  • Custom Agent A direct editor for people who want more control over prompts, models, knowledge, and draft setup before publishing.

When To Choose Business Agent

Choose Business Agent if you:

  • are creating your first agent
  • want the fastest path to a working assistant
  • prefer guided questions over manual setup
  • want KriyaOS to create and publish the first version for you

When To Choose Custom Agent

Choose Custom Agent if you:

  • want more control from the start
  • already know how you want the agent to behave
  • want to fine-tune prompts and settings before anything goes live
  • prefer to create a draft first and publish later

Create A Business Agent

The Business Agent path is the more guided option.

Step 1: Open The Guided Flow

From the Agents page, choose Business Agent.

Step 2: Fill In The First Setup Screen

You will be asked for:

  • your workspace
  • your agent name
  • your business type
  • your primary goal

This step is meant to help KriyaOS understand what kind of assistant you want to create.

Step 3: Teach The Agent

In the next step, you can give the agent business context by:

  • describing the business
  • uploading files
  • choosing core tools

If you upload files, they may still need time to finish processing after upload.

Step 4: Deploy The Agent

When you click Deploy Agent, KriyaOS creates the agent and publishes it as part of the guided flow.

That means this path is not draft-first in the same way as the Custom Agent path.

Step 5: Review The Live Result

After deployment, you land on a live/share screen. From there you can:

  • preview the agent
  • copy a share link
  • use the QR code
  • continue into integrations later

Create A Custom Agent

The Custom Agent path gives you more control and starts with a draft.

Step 1: Open The Editor

From the Agents page, choose Custom Agent.

Step 2: Fill In The Main Setup Areas

The editor includes:

  • Basic Information The name and description of the agent
  • Prompt Configuration The main instructions that shape how the agent behaves
  • Model Selection The AI model the agent uses
  • Knowledge Sources Any ready knowledge sources you want to connect
  • Advanced Settings Response and session settings

Step 3: Create The Draft

When you click Create Agent, KriyaOS creates the agent as a draft.

That means:

  • the agent exists
  • you can keep editing it
  • you can test it
  • it is not live yet

Step 4: Refine And Test

After creation, you stay in the agent editor. From there you can:

  • save draft changes
  • open the playground
  • publish when you are ready

What The Main Fields Mean

You do not need to understand every advanced setting right away. The most important fields are these:

Name

The name is how you will recognize the agent later.

Choose something clear and easy to identify.

Description

This is a short note for your own reference. It helps you remember what the agent is for.

System Prompt

This is the main instruction for the agent.

In plain language, it tells the agent:

  • who it is
  • what it should help with
  • how it should respond

If you only spend extra time on one field, make it this one.

Developer Instructions

This label sounds technical, but you can think of it as extra guidance for how the agent should operate.

If you are new, you can often leave this simple.

Guardrails And Constraints

These are the rules the agent should not break.

Use this area for boundaries, limits, and safety rules.

Primary Model

This is the main AI model the agent uses to answer.

Most new users can start with the default and change it later if needed.

Fallback Model

This is a backup model.

It is optional and mainly useful if you want an extra safety net.

Knowledge Sources

These are the files or information sources the agent can use when answering.

In the Custom Agent editor, you will only be able to select knowledge sources that are ready to use.

What Happens After You Create An Agent

What happens next depends on the path you chose.

After Business Agent Creation

Your agent is created and published as part of onboarding.

You can then:

  • preview it
  • share it
  • come back later to improve it

After Custom Agent Creation

Your agent starts as a draft.

You can then:

  • continue editing
  • test the draft in the playground
  • publish it when you are happy with it

Draft Vs Published

This is one of the most important product concepts to understand.

  • Draft means the version you are still working on
  • Published means the live version

The Custom Agent path creates a draft first.

The Business Agent path creates and publishes as part of onboarding.

Common Mistakes And Tips

  • If you want the fastest path, choose Business Agent. If you want more control, choose Custom Agent.
  • Do not confuse draft with live. A draft is not visible as the published version until you publish it.
  • If uploaded knowledge is still processing, the agent may not answer with that information right away.
  • The label Developer Instructions can sound more technical than it needs to be. Treat it as extra operating guidance.
  • In the guided onboarding flow, some labels may feel business-oriented rather than agent-oriented. That is normal because the flow is trying to help KriyaOS build the first version for you.

Next Step

If you have not already, it is also helpful to read What Is an Agent? so the draft and published steps feel more intuitive.