🤖Creating AI Games Without Coding

Overview

This document explains how to create game content in OVERDARE using AI. AI assists with various tasks such as level design, UI composition, and feature implementation, making the development process faster and more efficient.

This feature is currently in a pilot phase. Some functionalities may be limited or may produce results that differ from expectations.

Use Cases

The AI Agent can be used for the following tasks:

• Reviewing design documents and suggesting development directions

• Implementing game features (automatic script generation)

• Generating ideas for maps and level design

• Designing UI/UX structures

• Searching for and recommending assets

OVERDARE Games Created with AI

Here are examples of OVERDARE games created using AI.

Explore how maps, features, and UI were developed with AI assistance. Even simple ideas can be turned into fully playable games.

3v3 Escort Action Game

Object-Smashing Survival Game

Whack-a-Mole Action Game

Bumper Battle Game

Roguelite Action Game

Wave Defense Game

Top-Down Horror Shooter

Creeper Escape Game

Horror Escape Game

Bat Swing Simulator

Free Kick Soccer Game

Try It Yourself!

How to Use

Creating a Game with AI Agent

1. When you launch OVERDARE Studio, the OVERDARE AI Agent panel appears on the left side of the screen.

   
2. If the Windows protected your PC message appears, click More info, then click the Run anyway button. (No need to worry. This message appears because the application is not yet widely recognized or has a low download count.)

3. If a Windows security dialog requesting network access permission appears, click the Allow button. (No need to worry. This is a normal procedure required for the application to use network features.)

4. Click the Connect ChatGPT or More options button displayed in the OVERDARE AI Agent panel to connect an AI language model (LLM).

5. Enter your development request (prompt) in the OVERDARE AI Agent chat input field, then click the Send button.

6. Wait for the AI to complete the requested task.

7. Review the AI’s response and verify the generated results in Studio. (Always review and validate AI-generated results before use to ensure they accurately reflect your intended outcome.)

8. Continue refining the results by submitting additional modification requests.

By repeating this process, you can quickly turn your ideas into playable game content and verify the results in real time.

Add Object Information to an AI Agent Chat

After selecting an object in the Viewport or Level Browser, click the Add to Chat button to add the selected object to the AI Agent chat.

The added object is included as context in the chat input area, allowing you to request object-related tasks from the AI Agent.

Additionally, if the Studio Settings > AI Agent > Show Floating Panel on Viewport Selection option is disabled, the floating panel will not appear when selecting objects in the Viewport.

Show or Hide the AI Agent Panel

In Agent mode, you can click the AI Agent button on the top toolbar to show or hide the panel.

In Editor mode, use the AI Agent button in the View tab of the top toolbar to show or hide the panel.

Switching Between Agent and Editor Modes

You can switch Studio modes by clicking the AI Agent or Editor button in the upper-right corner of the screen.

• AI Agent Mode: A streamlined creation environment centered around the AI Agent panel, designed to help users create content with minimal setup and complexity.

• Editor Mode: An advanced creation environment that provides access to all Studio features and panels, making it ideal for experienced users and creators.

Tips

Write Clear Prompts

To achieve high-quality results, it is important to provide clear and specific instructions. Vague requests increase the likelihood of generating results that differ from your intent.

Script Implementation Troubleshooting Guide

If an error occurs during script execution or the result is different from expected, it is more effective to ask the AI to add logs rather than simply requesting a fix. Then, use those logs to analyze and resolve the issue.

First, ask the AI to insert logs into the main execution flow. Then, share the resulting logs back with the AI so it can identify where the issue occurred and analyze the root cause.

Example

1. (When an issue occurs) “It seems like ~~ is not working. Please add logs step by step.”

2. Run a play test and copy the generated logs

3. “Analyze the cause of the issue based on these logs and fix it.”

4. Repeat steps 1–3 until the issue is resolved

Session Management

As conversations become longer, the context referenced by the AI can accumulate excessively, which may lead to a decrease in result quality. In particular, when a single session continues for a long time, the AI may generate responses that differ from your original intent.

If the results seem incorrect or repeatedly produce unexpected outputs, it is recommended to start a new session and try again.

In the following cases, click the New conversation button to start a new chat:

• When switching to a completely different topic (e.g., from map creation → UI implementation)

• When the answers become increasingly inaccurate for the same question (e.g., initially correct responses → repeated incorrect code or explanations)

• When the responses become overly long and drift away from the intended answer (e.g., you wanted a simple solution, but unnecessary details keep being added)

Notes

When you start a new session, previous conversation history is not carried over automatically. Therefore, if you want to continue working on a previous task, it is recommended to summarize the current conversation before switching sessions.

You can ask the AI to summarize the work so far, then copy and paste that summary into the new session to maintain context. This approach helps reduce errors caused by unnecessary context accumulation and keeps the workflow more stable.

Make Active Use of Result Summaries

The AI provides summaries or explanations along with the results. These summaries include key information needed to understand and refine the output.

For example, you can find the following information:

• Description of the generated structure (e.g., how the feature is organized)

• Key parameters (e.g., speed, size, conditions, and other core values)

• Adjustable points (e.g., elements that will affect the result when modified)

If the result is not what you expected, using the summary as a reference when requesting changes will help you improve the result more quickly and accurately.

Advanced Usage

Using AGENTS.md

By creating an AGENTS.md file at the root of your project folder, you can clearly define the standards and rules that the AI should follow during its tasks.

With this file, the AI can understand your project’s workflow and guidelines without requiring repeated explanations.

Changes to the md file are applied when the Studio is launched.

Example File

Add the provided AGENTS.md file to your project folder, then modify its contents to fit your project.

This file may include rules such as:

• Organizing request details and results for each task

• Creating documentation outputs after planning/development work

• Updating existing documents instead of creating new ones when they are closely related or continuous

Extending Project Rules

You can also add additional rules as needed, such as:

• Project structure design guidelines

• Documentation writing and management standards

• Methods for organizing results

• Task units and workflow guidelines

• Reusability and modification standards

Using SKILL.md

If you add Skill folders under the project's /.diligent/skills directory, the AI can reference task-specific instructions and perform work in a more consistent way.

How to Use

Create a folder for each Skill inside /.diligent/skills, then add a SKILL.md file inside each folder.

Write the name and description fields in the metadata section of SKILL.md.

• name should match the Skill folder name exactly.

• Do not use underscores (_) in file names or folder names.

• description should clearly explain in a single sentence when the AI should use this Skill.

Changes to the md file are applied when the Studio is launched.

How OVERDARE AI Agent Works

OVERDARE’s AI features are powered by an Agent system, which connects an AI language model (LLM) such as ChatGPT with Studio tools (MCP) to perform actual tasks.

• LLM (Large Language Model): Understands user requests and generates a plan &#xNAN;(e.g., models like ChatGPT)

• Agent: Acts as an executor that translates requests into actual operations

• MCP Studio Tool: An interface that directly executes functions within the Studio

In summary, tasks are processed through the following flow:

User Input → LLM Interpretation → Agent Execution → Studio Tool Operation