Unity Mcp

Open source alternative to Unity and Coplay

15.4K stars
MIT
Active this week
Last commit 5 days ago

Last verified Oct 25, 2026

Connect AI assistants to the Unity Editor via the Model Context Protocol, exposing 50 tools to manage assets, scenes, scripts, and game workflows.

Visit website GitHub15.4K Deploy on Hostinger

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who Unity Mcp is for#

Game developers automating repetitive Editor tasks

Drive scene setup, asset organization, and prefab configuration with natural language through Claude, Cursor, or any MCP client. Instead of writing one-off Editor scripts, describe the change to your AI assistant and have it execute directly inside the Unity Editor through the 50-tool surface.

Skip if:

Skip if your repetitive tasks happen outside the Unity Editor process (build pipelines, asset bundle generation, CI/CD steps), where Unity MCP's Editor-scoped tools add no value.

Technical artists working on VFX, animation, or UI systems

Activate the VFX, animation, or UI tool groups on demand to expose a focused set of tools for your current domain. Per-session visibility keeps the prompt lean and routing accurate when your AI client is working on one specific area of a large project.

Skip if:

Skip if your primary bottleneck is runtime behavior rather than Editor-side authoring. Unity MCP only exposes Editor APIs; it does not expose the running game's state.

Teams running shared AI-assisted development pipelines

HTTP transport and multi-instance routing let a team run one shared MCP server and route different AI clients to separate Unity Editor instances. Remote server authentication support means the Python server can run on a shared development machine accessible to multiple developers.

Skip if:

Skip if the team's bottleneck is code generation rather than Editor control. Unity MCP's value is the Editor integration layer; it does not replace a code-generation tool.

AI researchers studying LLM-driven 3D authoring

The project was cited at ACM SA Technical Communications 2025 (Wu and Barnett, MCP-Unity: Protocol-Driven Framework for Interactive 3D Authoring). Unity MCP's structured MCP tool surface, documented three-layer architecture, and open MIT codebase make it a platform for studying LLM-Editor interaction at the protocol level.

Skip if:

Skip if your research requires a runtime game-state loop rather than Editor-side authoring; Unity MCP does not expose the running game's state to MCP clients.

the problem

The problem it solves#

Integrating AI assistants into Unity game development requires either writing custom Editor scripts to expose Unity APIs or relying on proprietary plugins that charge subscription fees and send project data to external servers. Most Unity developers who use Claude, Cursor, or Copilot for code generation still copy-paste generated code into the Editor manually, losing all context their AI has about the current scene state.

The gap is that AI clients have no standard way to read or modify what is happening inside the Unity Editor. Without a bridge layer, developers context-switch constantly between their AI assistant and the Editor, manually translating natural-language instructions into Editor API calls one step at a time.

how Unity Mcp solves it

How it solves it#

50 MCP tools across 10 tool groups

Exposes scenes, GameObjects, assets, scripts, prefabs, materials, animation, VFX, UI, testing, and ProBuilder through focused MCP entrypoints. Tool groups activate per session so only the relevant tools appear in your AI client's context, reducing prompt size and improving routing accuracy on large projects.

Three-layer architecture: client, Python server, C# plugin

The Python server (FastMCP + WebSocket hub) sits between your MCP client and the Unity Editor, handling per-session routing via client_id, active instance tracking, and hot reload. The C# plugin runs on Unity's main thread and executes commands through standard Editor APIs, returning structured payloads the MCP client can reason over.

HTTP and stdio transport options

Supports stdio for single-client setups like Claude Desktop and HTTP for multi-agent, remote-hosted, and shared workflows. Transport mode is auto-detected and auto-configured at startup, so switching between local and remote deployment requires no manual JSON editing.

Multiple Unity instances, single MCP session

Aim one MCP session at any of several simultaneously running Unity Editors. Per-call routing lets cross-project prompts target different Editor instances in the same conversation, while session isolation ensures separate MCP clients do not interfere with each other.

25 read-only MCP state resources

Exposes 25 read-only resources for querying Editor state: scene hierarchy, asset metadata, script contents, and component configurations. Resources let an AI client inspect the project before executing write operations, improving accuracy on multi-step authoring tasks.

Extensible custom tool registration

Register new MCP tools with a single C# attribute. The Python server discovers them automatically through reflection-based dispatch. Custom tools can be project-scoped or global, and no Python-side changes or server restarts are needed when adding new domains.

strengths · trade-offs

Strengths and trade-offs#

Strengths

MIT license, no telemetry, no account requiredThe full toolset is MIT-licensed and runs on your own machine with no sign-up and no telemetry enabled by default. Unlike Unity's own AI features or commercial plugins like Coplay, you retain full ownership of project data, and there is no subscription to cancel if the vendor changes pricing.
24 MCP clients auto-configured on installThe installer detects and configures 24 MCP clients including Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Cline, Codex, and Gemini CLI through Window > MCP for Unity > Configure All Detected Clients. Most users need no manual JSON editing to get their preferred AI client connected.
Actively maintained with an academic citation15,000+ GitHub stars and regular releases (v10.3.0 in October 2026). The project was cited at ACM SIGGRAPH Asia Technical Communications 2025 in Wu and Barnett's MCP-Unity: Protocol-Driven Framework for Interactive 3D Authoring. Maintained by Aura, the team also building commercial Unity and Unreal AI tools.
Unity version range covers most production projectsSupports Unity 2021.3 LTS through Unity 6.x LTS, covering the range most production projects target. The README provides version pinning instructions for stability, and the project's own versioning (v10.x) tracks Unity's major LTS release cycle.

Trade-offs

-Requires Python 3.10+ as a separate runtimeThe Python server (run via uv) is an additional runtime dependency alongside Unity. Developers without Python in their environment need to install it before the server starts. On locked-down corporate machines, managed CI environments, or shared lab computers, this can block initial setup without admin access.
-113 open GitHub issues at time of writingWith 113 open issues, some edge cases and compatibility problems remain unresolved. Users on Unity versions at the boundaries of the supported range (2021.3 LTS, Unity 6 LTS) may encounter unpatched bugs before the next release. High star counts attract a large user base that surfaces more issues than smaller projects.
-Not affiliated with Unity TechnologiesUnity MCP is maintained by Aura, not Unity Technologies. If Unity's Editor API surface changes significantly in a future release, the C# plugin may require updates before it functions again. The README includes an explicit disclaimer of non-affiliation, and there is no guarantee of Unity-side compatibility testing.

versus alternatives

Unity Mcp vs alternatives#

Unity MCP vs Unity's Built-in AI Features

Unity Technologies has progressively added AI features to the Unity Editor, including Unity AI (code generation and scene assistance) and integrations with GitHub Copilot. These are proprietary, require a Unity subscription tier that includes the AI features, and route commands through Unity's cloud infrastructure.

Feature	Unity MCP	Unity AI / Copilot Integration
License	MIT	Proprietary
Self-hosting	Yes (Python server + C# plugin)	No
AI client compatibility	Any MCP client (24 auto-configured)	GitHub Copilot, Unity-specific clients
Telemetry	None by default	Cloud-dependent
Cost	Free	Subscription-based

Unity MCP is the better choice when you want to use your own AI client (Claude, Cursor, local LLMs) without being locked into Unity's preferred AI vendors. It is also the right call when data privacy matters: all commands route through your own Python server with no external relay. Unity's built-in AI features have the advantage of tighter roadmap integration and official support from Unity Technologies.

Unity MCP vs Coplay

Coplay is a premium AI assistant for Unity and Unreal, made by Aura, the same team that maintains Unity MCP. Coplay targets professional studios that want a polished, commercially supported tool rather than an open source integration layer they manage themselves.

Feature	Unity MCP	Coplay
License	MIT	Proprietary / Paid
Price	Free	Commercial subscription
Self-hosting	Yes	No
Support	Community / GitHub Issues	Commercial support
AI client	Any MCP-compatible client	Coplay client

Unity MCP is the right choice for developers who want to bring their own AI client and do not need commercial support or a managed onboarding experience. Coplay is worth evaluating if your studio needs a fully integrated AI experience in Unity with dedicated vendor support and is willing to pay for it. The two tools come from the same team, so the underlying Editor integration quality is comparable; the difference is ownership model and support level.

install · quick start

Quick start#

bash

Install the C# plugin via Unity Package Manager using the git URL below, or clone the repository to add the package from disk.
```bash
git clone https://github.com/CoplayDev/unity-mcp.git
```

tech stack · detected from GitHub

What it's built on#

Languages: C#Python
Frameworks: FastAPIReact

frequently asked

FAQ#

Does Unity MCP work with Claude, Cursor, Copilot, and other AI clients?

Yes. Unity MCP supports any MCP-compatible client. The installer auto-configures 24 clients including Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Cline, Codex, and Gemini CLI. Clients using stdio get single-client setups; HTTP transport supports multi-agent and shared server workflows.

What Unity versions does Unity MCP support?

Unity 2021.3 LTS through Unity 6.x LTS. The project's own versioning (currently v10.3.0) tracks Unity's major releases. Pin a specific version in the Package Manager git URL with a tag such as #v10.0.0 if you need a stable installation tied to a known release.

Is there a hosted or managed version of Unity MCP?

No. Unity MCP is a self-hosted tool; the Python server and C# plugin run entirely on your machine or your own infrastructure. There is no managed cloud service and no relay between your AI client and the Unity Editor. The README confirms that no telemetry is enabled by default.

Can I add custom tools to Unity MCP beyond the built-in 50?

Yes. Add a C# attribute to register a new tool in the MCP surface. The Python server discovers it automatically through reflection-based dispatch. Custom tools can be project-scoped or global and require no changes to the Python server or any restarts after registration.

What is the relationship between Unity MCP and Coplay?

Coplay is a commercial AI assistant for Unity and Unreal made by Aura, the same team that maintains Unity MCP. Unity MCP is the free, MIT-licensed open source integration layer; Coplay is Aura's premium product. They address the same general use case (AI assistance in Unity) but at different price and support levels.

also worth a look

Similar open-source tools#

Godot Engine

Alternative toUnity

Free open source game engine for 2D and 3D development

118,973 GitHub stars
Licence MIT

Hallmark

InDesign & Creative

Design skill for coding agents with 48 themes, real photos and lint.

32,161 GitHub stars
Licence MIT

Openfilm

Alternative toRemotion

AI video agent for coding agents, with a local editor and no account.

2,601 GitHub stars
Licence MIT

SOMA

Alternative toZoo.dev

Parametric 3D modeling via CLI agent: describe, generate, iterate.

286 GitHub stars
Licence GPL-3.0

OpenCreator

Alternative toHeyGen

Local AI workspace for video, image, voice, and translation

13,875 GitHub stars
Licence Apache-2.0

Artcraft

Alternative toRunway

Open source AI studio for artists who want real control

35,664 GitHub stars
Licence Apache-2.0

by category

Open source tools in these categories#

478

Run Unity Mcp on your own server

Self-host Unity Mcp on a Hostinger VPS.

Deploy on Hostinger

Featured

Omnara

Alternative toClaude Managed Agents

Open-source agent deployment API. Self-host or use Omnara Cloud.

3,235 GitHub stars
Licence Apache-2.0
#AI Agents

Inbox Zero

Alternative toSuperhuman

AI email assistant that drafts replies and clears your inbox

12,787 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

5,406 GitHub stars
Licence Apache-2.0
#AI Agents

Feature your project

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Unity MCP vs Unity's Built-in AI Features

Feature	Unity MCP	Unity AI / Copilot Integration
License	MIT	Proprietary
Self-hosting	Yes (Python server + C# plugin)	No
AI client compatibility	Any MCP client (24 auto-configured)	GitHub Copilot, Unity-specific clients
Telemetry	None by default	Cloud-dependent
Cost	Free	Subscription-based

Unity MCP vs Coplay

Feature	Unity MCP	Coplay
License	MIT	Proprietary / Paid
Price	Free	Commercial subscription
Self-hosting	Yes	No
Support	Community / GitHub Issues	Commercial support
AI client	Any MCP-compatible client	Coplay client

Who Unity Mcp is for#

Game developers automating repetitive Editor tasks

Skip if:

Technical artists working on VFX, animation, or UI systems

Skip if:

Teams running shared AI-assisted development pipelines

Skip if:

AI researchers studying LLM-driven 3D authoring

Skip if:

The problem it solves#

How it solves it#

50 MCP tools across 10 tool groups

Three-layer architecture: client, Python server, C# plugin

HTTP and stdio transport options

Multiple Unity instances, single MCP session

25 read-only MCP state resources

Extensible custom tool registration

Strengths and trade-offs#

Strengths

Trade-offs

Unity Mcp vs alternatives#

Unity MCP vs Unity's Built-in AI Features

Unity MCP vs Coplay

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

Design & Creative

Find open source alternatives to paid software

Who Unity Mcp is for#

Game developers automating repetitive Editor tasks

Skip if:

Technical artists working on VFX, animation, or UI systems

Skip if:

Teams running shared AI-assisted development pipelines

Skip if:

AI researchers studying LLM-driven 3D authoring

Skip if:

The problem it solves#

How it solves it#

50 MCP tools across 10 tool groups

Three-layer architecture: client, Python server, C# plugin

HTTP and stdio transport options

Multiple Unity instances, single MCP session

25 read-only MCP state resources

Extensible custom tool registration

Strengths and trade-offs#

Strengths

Trade-offs

Unity Mcp vs alternatives#

Unity MCP vs Unity's Built-in AI Features

Unity MCP vs Coplay

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

Design & Creative

Find open source alternatives to paid software