COMING SOON Release candidate · Not yet available on the Unity Asset Store

AI ASSISTANCE FOR UNITY GAME MAKERS

Give your AI
something
useful to do.

Bring an AI assistant into Unity, the software used to make games. Ask it to inspect your project or help make changes, choose what it can access, and review its work.

Free release planned. Hosted AI services require your own account and may charge usage fees. No model or inference hosting included.

KANBAN GRAPHILLUSTRATED WORKFLOW
EXAMPLE / ADD A QUEST SYSTEMA shared goal. Work with clear roles.
Multiple agents contributing to a quest systemPlan the feature branches into quest logic and quest interface. Both feed integration, which leads to review. Arrows show the order of work. This is an illustrative workflow. Plan the featurePlanner · CompleteQuest logicDeveloper · WorkingQuest interfaceUI agent · WorkingConnect & checkIntegrator · QueuedReview the resultReviewer · QueuedBRING BOTH PARTS TOGETHER
↳
Different agents. One connected feature.

A developer handles quest logic while another agent builds the interface. Their work comes together for integration and review.

Who is it for? People making games or tools in Unity. Some Unity experience is needed; creating your own integrations requires C#.

What does it do? Connects an AI assistant to project tools, with a Kanban board to organize tasks and a graph to see how they fit together. You direct the work and check the results.

1,033 / 1,033

Offline EditMode tests passed

34 / 34

Asset Store validator checks passed

2022.3 LTS

Tested on Windows Editor · 2022.3.30f1

Candidate validation, 28 October 2026.
Not an Asset Store approval or a live-provider guarantee.

01 / ORGANIZE THE WORK

A task board.
A clearer picture.

Kanban puts work on cards so you can follow what needs doing, what an assistant is working on, and what needs review.

KANBAN / FOLLOW PROGRESS

Keep track of the work.

Give a task a clear goal, assign an assistant, and follow its progress on the board. Open a card to inspect the details and decide what needs attention.

  • See tasks and their current status.
  • See which assistant is working or waiting.
  • Review results and request changes when needed.
Read about the task board ↗
KANBAN GRAPH / FOLLOW CONNECTIONS

Understand what is waiting on what.

The same tasks appear as connected circles. The lines show parent and child tasks, dependencies, blockers and other associations.

  • Spot the task holding up other work.
  • Pan, zoom and arrange the view to explore it.
  • Double-click a task to open its details.
Explore Kanban Graph ↗
MULTI-AGENT WORKFLOWS

Give each assistant a role.

Work with multiple AI agents on the same project. Assign tasks to different assistants, choose their models and permitted tools, and follow their work together on the Kanban board.

Task dependencies help keep the work in order. You can see who is working, who is waiting, and what needs your review.

Read about multiple agents ↗
FALLBACK PROVIDERS

Have a backup ready.

Configure an ordered list of backup providers. If a request hits a supported failure, such as a timeout, rate limit or unavailable service, AIRuntime can try the next configured provider.

You choose the backups. Each needs its own working configuration, and hosted services may have separate costs.

Read about provider fallback ↗

HOW THE EXAMPLE WORKS

Give each agent a clear piece of the job.

For a quest system, one agent can work on objectives and completion logic while another handles the on-screen quest list. Integration brings those pieces together, and review checks the result.

The graph makes those relationships visible: who owns each task, what can move forward, and what is waiting for another piece.

Two views of the same work. Moving a circle in the graph changes its layout, not the task’s dependencies or execution order.

02 / TWO WAYS TO BUILD

Use it in Unity.
Build on it with code.

01 / FOR PROGRAMMERS

Build AI features of your own.

Give your own application an AI conversation and a set of tools it can use. Programmers can extend AIRuntime with C#.

  • Keep work running without blocking, and cancel when needed
  • Define the tools the AI can use
  • Set permissions and inspect the results
  • Handle service failures with retries and backup providers
See the integration shape ↗
02 / INSIDE UNITY

See the work as it happens.

Chat with an assistant inside Unity, organize its work on a task board, and see what it is doing.

  • Connect your chosen AI service and start a conversation
  • Choose what each assistant is allowed to do
  • Follow the work on a task board
  • Review generated changes before relying on them
Find the Editor tools ↗

CONNECT YOUR CHOSEN AI SERVICE

One framework.
Your choice of model service.

OpenAIOpenRouterNous PortalNVIDIA DeveloperLocal AI (Ollama)

Use a hosted AI service, or connect to a local model running on your own computer through Ollama. You supply the model and the hardware for local AI. Adapters are included; accounts, models and hosting are separate. Tool support and options vary by model and service.

03 / YOUR FIRST TEN MINUTES

Start small.
See a real result.

Read the getting-started guide ↗
  1. 01

    Add the required Unity packages.

    Newtonsoft JSON, Unity Test Framework and Unity UI. Exact versions and built-in module requirements are in the guide.

  2. 02

    Import AIRuntime.

    The package lives under one Assets/AIRuntime folder, with source, documentation and examples together.

  3. 03

    Try the example without an AI account.

    Open the included scene or launch the example from Window → AIRuntime → Examples. No credentials needed.

BUILT FOR DEVELOPERS, WITH CLEAR BOUNDARIES

You choose what the agent can do.

Enabled tools can read or change project files and assets. Hosted requests can contain prompts, conversation history and tool results. Review your permissions, provider terms and generated changes.

Read the data and credential notes ↗

Permission policies are not an operating-system sandbox. Exported profiles can contain plaintext API keys. Keep them outside build assets and version control.

HOW THIS PROJECT WAS MADE

100% made by AI.
With real human guidance.

AIRuntime is an AI-made project, directed by a real person. AI does the creation; human guidance sets the goals, shapes the direction and decides what to keep.

This describes how AIRuntime was made. Unity, third-party dependencies and connected AI services are separate products with their own authors and licenses.

04 / MADE BY VOLTRON

Get unstuck.
Get back to building.

Questions, bugs or a rough edge? Send the Unity version, operating system, reproduction steps and a redacted error log.

mjdust.defender@gmail.com
RELEASE STATUS

1.0.0 candidate

Prepared for review. Asset Store submission and public availability are still pending.

Ask about the release ↗Free release planned · Asset Store submission pending