Unity’s AI tools are a suite of tools built into the Unity Editor for developers using Unity 6 or later. They aren’t a separate product or brand — the tools are part of Unity itself. They include an in-editor AI assistant (a project-aware AI agent that runs inside the Editor), AI generators (which produce placeholder materials, sounds, cubemaps, and other dev-stage assets), an AI gateway (which lets developers bring third-party agents like Claude, Codex, or Gemini into Unity), and the official Unity MCP server (which lets external tools and IDEs control the Unity Editor).
This page covers:
• The four AI tools in Unity
• When to use them (and when not to)
• How they compare to general AI coding tools
• AI asset tagging and traceability
• Data handling
• Getting started
The four AI tools in Unity
These are Unity’s agentic AI tools that run inside the Unity Editor. There are four components.
The in-editor AI assistant is a project-aware AI agent that runs as a panel inside the Editor. It has direct context of the project’s scene hierarchy, settings, packages, code, asset library, platform targets, and serialized Inspector data — context a general-purpose LLM doesn’t have.
AI generators produce placeholder cubemaps, materials, sounds, 3D models, and similar dev-stage assets directly from the Editor. Every generated asset is tagged with metadata identifying it as AI-generated, so it can be audited and replaced before shipping.
The AI gateway lets developers connect verified third-party AI agents, including Claude, Codex, and Gemini, directly to the Unity Editor. The agent uses Unity’s project context without consuming Unity AI Credits; it runs on the developer’s existing subscription with that provider.
The official Unity MCP server lets developers control the Unity Editor from external tools and IDEs. This is the bring-your-own-tool surface, useful for developers who already work primarily in Cursor, Claude Code, or another IDE-based workflow and want Unity to participate.
Unity’s AI tools are distinct from Unity Sentis, which is the runtime engine for executing neural network models inside shipped games. Unity’s AI tools are for the developer’s authoring workflow; Sentis is for the player’s runtime experience.
They also aren’t a way to prompt a finished game into existence. This is coding and authoring assistance. A developer still designs the game, makes the architectural decisions, and ships it; the AI tools help with the work along the way.
How Unity’s AI tools fit into your project
Each of the four components addresses a different part of a developer’s workflow.
The in-editor AI assistant opens as a panel inside the Editor. Most developers reach for it for in-line troubleshooting, refactoring, and one-off agentic tasks — resolving a console error, generating a wired-up GameObject from a prompt, refining an animation transition, or building a small custom Editor tool.
AI generators sit alongside the Editor’s asset import pipeline. Output drops into the project’s asset folder with metadata identifying it as AI-generated, ready to be referenced like any other asset. The intended use is dev-stage placeholders, such as a credible cubemap or texture stand-in while final art is in production.
The AI gateway is configured in Project Settings. Once set up, third-party agents can use Unity’s project context the same way the in-editor assistant does. A team paying for Claude Code can keep that subscription and have it work better inside Unity.
The Unity MCP server is the outside-in control plane. It lets an external IDE, a CLI agent, or another MCP-aware tool drive the Unity Editor. For developers who already work primarily in another tool, the Unity MCP server brings Unity into that workflow rather than asking them to switch.
Why an Editor-native AI assistant is different from a general-purpose AI coding tool
A general-purpose AI coding tool sees what you paste into it, plus (sometimes) the open file in your editor. The in-editor AI assistant sees the live state of the Unity project: scene hierarchy, settings, packages, code, asset library, platform targets, and serialized Inspector data. It can read what’s actually in the scene, not just what’s in a file.
That difference matters because most Unity work isn’t expressible in source files alone. A GameObject’s behavior depends on the components attached to it, the values set in the Inspector, the scene it’s instantiated in, and the platform target the project is building for. A coding agent that can’t see those isn’t equipped to make changes that hold up when you press Play.
The in-editor AI assistant has three interaction modes:
Ask Mode is conversational: guidance, troubleshooting, explanations. It’s the mode developers reach for first.
Agent Mode is multi-step task execution — the mode that turns the Editor itself into an action surface, creating GameObjects, wiring components, and modifying scenes.
Plan Mode helps you design before you build. Built for game development, it helps you turn loose ideas into a clear implementation plan. When it’s time to execute, it follows long end-to-end instructions, including full Game Design Documents, rather than skipping steps or calling things done before they are.
The gateway model means Unity’s AI tools don’t ask a developer to abandon the IDE they already use. A team paying for Cursor can plug it into Unity through the AI gateway, and it gets the same project context the in-editor assistant has.
AI generators tag every asset they produce with metadata identifying it as AI-generated. That tag answers a specific question studios ask before shipping: where are the AI-generated assets in this project, and have they been reviewed? The metadata makes that question answerable.
When to use Unity’s AI tools (and when not to)
Unity’s AI tools are the right choice when:
• Your project is on Unity 6 or later. (This is a hard requirement — the AI tools don’t run on earlier versions.)
• You’re doing in-Editor work — scene composition, animation tuning, GameObject wiring, custom Editor tools, debugging.
• You’re prototyping or iterating, and placeholder assets help validate a mechanic faster than waiting on final art.
• You already pay for Claude Code, Cursor, or another agent, and want it to work better inside Unity. (Use the AI gateway.)
• You want to use a Unity-official tool rather than a third-party Unity-specific one — for support, alignment with Unity’s roadmap, or procurement reasons.
They’re less suitable when:
• You’re hoping to generate a finished game from a prompt. This is coding and authoring assistance, not a game generator. A developer who can’t write C# can still benefit — the assistant can scaffold projects and explain code as it generates — but learning Unity’s basics is still part of the path.
• You need a senior engineer’s judgment on architecture. The AI tools are good at execution and explanation. Decisions like “should this project use ECS or GameObjects” still belong to a person who understands the trade-offs.
• You’re working in a game engine other than Unity. These tools are for Unity.
• Your work isn’t Unity work at all. They don’t help with non-Unity tasks.
How Unity’s AI tools compare to other AI tools
Unity’s AI tools most closely compare to Bezi, a third-party Unity-specific tool, and to general AI coding tools like Cursor, Claude Code, and Copilot. Their main advantages over both are that they’re built natively into the Editor, made by Unity, and have direct access to your project’s full state: scene hierarchy, Inspector values, serialized data, and platform targets.
Unity’s AI tools vs. Bezi
Bezi is the closest alternative for Unity-specific work. It runs as a plugin alongside the Editor rather than inside it, so its scene awareness is more limited. It’s code-focused and doesn’t natively drive the Editor for scene or spatial work the way Agent Mode does. Play Mode verification requires more iterations. It has its own 3D asset generation pipeline. MCP support is on the roadmap. Pricing runs $20, $60, or $200 per month, with custom pricing options for Enterprise plans.
Unity’s AI tools vs. general AI coding tools
Cursor, Claude Code, and Copilot run in your IDE or terminal. They see your open files and can read project files through the Unity MCP server, but they don’t natively understand Unity’s serialized state or what’s actually in your scene. They can drive the Editor through the MCP server, but it’s indirect. They don’t verify changes in Play Mode and don’t generate game assets. These tools aren’t really alternatives to Unity’s AI tools — they’re complements. The AI gateway lets you connect your existing Cursor or Claude Code subscription to Unity’s project context without spending Unity AI Credits.
Competitive claims about Bezi are based on publicly available documentation. Comparison capabilities change quickly. Last updated: July 2026.
Tagging and traceability: How Unity handles AI-generated assets
Every asset produced by a Generator is tagged with metadata identifying it as AI-generated. The tag is set at generation time and travels with the asset.
That metadata makes generated assets findable at audit time. A developer or compliance reviewer can search the project for AI-generated assets, review them, and replace any that aren't intended to ship. Unity recommends treating Generator output as placeholder content by default. Credible enough to validate a mechanic, but intended to be replaced before ship.
First-party Generators use filters that reduce the risk of producing copyrighted material, but no filter is foolproof. Third-party models brought in through the AI Gateway are subject to the policies of those providers; Unity doesn't add a separate filter on top.
For studios shipping commercially, the workflow is straightforward: generate placeholders to keep iteration fast, find the generated assets via metadata before ship, replace the ones that need to be replaced. The metadata makes the audit step tractable.
How Unity handles your data
By default, data from a developer's use of Unity’s AI tools isn't used to train AI models.
Developers who want to share data for training can opt in via the Unity Dashboard. Opt-in is per organization, not per seat.
Third-party agents brought in through the AI Gateway are subject to the data policies of those providers. If a team uses Claude Code through the Gateway, Claude Code's data handling applies to that traffic.
Pricing, access, and getting started
How credits work, who gets access at each tier, and how to start using Unity’s AI tools today.
How Credits Work
Unity AI Credits meter usage of the in-editor AI Assistant and AI Generators. Credits reset monthly and don't roll over. For the longer explanation of the credit model, see the Unity AI Credits guide.
Access Tiers
Access works three ways, depending on your Unity license and whether you want to use Unity's own agent or bring your own:
• Personal Users: Can try Unity’s AI tools with a 14-day trial that includes 1,000 credits (credit card required; converts to a $10/month Personal subscription unless cancelled).
• Pro, Enterprise, and Industry Subscribers: Already have the AI tools included with their seats. A monthly credit allotment covers Assistant and Generator use, and these tiers include access to the AI Gateway and the official Unity MCP Server. To turn it on, install the Assistant package from inside the Editor.
For live tier pricing, see the pricing page.
Getting Started with Unity’s AI tools
1. Install Unity 6. Download Unity 6 or later from the release archive or the Unity Hub. Earlier versions don’t run the AI tools.
2. Install the Assistant package. In the Editor, click the AI button and install the Assistant package.
3. Link the project to Unity Cloud. Unity’s AI tools require the project to be linked to Unity Cloud.
For more detailed instructions on getting started view the complete getting started guide.