Unity 2022.3.22f1+ · One-time purchase

Connect AI directly to Unity Editor.

Escape the screenshot grind — your LLM edits scenes, runs scripts, and checks the screen directly over file exchange. No server, no port, no account required.

View on Unity Asset Store

The Value We Deliver

Everything an AI needs to see, act, and verify — built entirely on file exchange.

01

Operate Unity Editor in natural language

From scene editing and script execution to GameObject manipulation, camera control, Play Mode control, and running tests — the LLM calls 20 tools across 6 groups (editor / scripting / scene / camera / play_mode / testing) directly.

02

Run any C# code on the spot

The core tool, execute_dynamic_script, runs arbitrary C# code instantly inside Unity Editor via in-process Roslyn (4.9.2). Requests that don't fit a fixed operation API can still be handled by writing code on the fly.

03

Escape the screenshot grind

The screenshot tool sends the Scene View, Game View, an arbitrary camera angle, or a rendered result to the LLM as an image. UI and 3D element annotation is supported too, so you never have to manually attach a screenshot to say "look at the screen now" again.

Architecture

No server, no port. The LLM and Unity are connected purely through file exchange.

How file transport works

The LLM coding agent writes a request to `<Project>/EditorEye/inbox/{id}.json`. Unity Editor watches this folder, executes the request as soon as it's detected, and writes the result back to `<Project>/EditorEye/outbox/{id}.json`. The agent simply reads that file to get the result.

  • No port listening, no TCP/WebSocket connection, no external server
  • No background process — it works whenever Unity Editor is open
  • Check transport status (pending requests, etc.) via Tools → Editor Eye StandAlone → Status Window
  • The core of every operation is execute_dynamic_script, which runs arbitrary C# code instantly via in-process Roslyn 4.9.2

Image architecture (checking Scene View / Game View)

The screenshot tool captures the Scene View, Game View, an arbitrary camera state, or a rendered result as an image, selected via the Source parameter. It can also return the image with UI (Canvas/UGUI) or 3D (collider detection via a physics raycast) element annotations, so the LLM can check "what's on screen right now" without you manually attaching a screenshot each time.

  • game_view Captures the Game View as-is
  • scene_view Captures the Scene View's current viewpoint (default)
  • custom_camera Temporarily switches to a specified camera state and captures from there
  • rendering Captures the Play Mode render output (supports UI annotation)

Up and running in 3 steps

Just import the .unitypackage. No config file editing required.

  1. 01

    Purchase and import from Unity Asset Store

    Buy Editor Eye StandAlone on the Unity Asset Store and import it into your project via Package Manager.

  2. 02

    Instruct your AI coding agent

    Just talk to an agent like Claude Code: "check this GameObject in Unity" or "make the material red." The agent writes a request into `EditorEye/inbox/`.

  3. 03

    Unity detects, executes, and responds automatically

    Unity Editor detects the request, executes it, and writes the result back to `EditorEye/outbox/`. There's no port and no background process — it works as long as Unity is open.

Installation Guide

Install the package, add its one dependency if prompted, then install the AI assistant documentation.

  1. 1.Import the package

    Import the .unitypackage into your Unity project. It installs to Assets/EditorEye.app/ — a normal Assets folder, not a UPM package.

  2. 2.Add the required dependency if prompted

    There is one required dependency: com.unity.nuget.newtonsoft-json. If it is not already in your project, a small setup window appears automatically the first time the Editor loads with this package present. Use its button to open Package Manager, add the package there, then let Unity recompile.

  3. 3.Install AI assistant documentation

    Open Tools / Editor Eye StandAlone / Setup Window and click Install AI Assistant Docs. Select the agents you use. The protocol document is written to EditorEye/AGENTS.md, and a short pointer is appended to the relevant files.

    • CLAUDE.md (Claude Code)
    • AGENTS.md (Codex and other agents)
    • GEMINI.md (Gemini CLI)
    • .cursor/rules/editor-eye.mdc (Cursor)

    Existing files are appended to, never overwritten. Files that already contain the pointer are left untouched, so running the installer again is safe.

  4. 4.Ask your agent to do something in Unity

    There is no config file to edit, no client to restart, no account, no API key, and no license or trial gate. This package has a single edition.

    The Setup Window also suggests adding EditorEye/ to your .gitignore. That folder holds transient request and response files; the installer does not edit .gitignore for you.

Frequently asked questions

What does Editor Eye StandAlone do? +

It's a Unity extension that lets LLM coding agents like Claude Code operate Unity Editor directly through file exchange. Scene editing, script execution, GameObject manipulation, and screenshots can all be driven by instructing the LLM.

What's the pricing model? +

It's a one-time purchase through the Unity Asset Store (a single SKU). There is no subscription, no license-key authentication, and no trial expiration.

Which LLM clients can I use? +

Any coding agent capable of reading and writing files (Claude Code, etc.) works. No connection setup to a specific MCP server is required.

Which Unity versions are supported? +

Unity 2022.3.22f1 or later.

Give your AI coding agent Unity's eyes and hands.

One-time purchase, no account required. Get started right away from the Unity Asset Store.

View on Unity Asset Store