Editor Eye : AI Agent Bridge & Safe Editor Automation

Plenty of tools connect AI to Unity.

But are you satisfied with the experience?

EES makes AI and Unity work better together.

Why does EES cut AI token usage by 56.4%?

Because every EES response is written like a Senior Unity Engineer coaching your AI on the exact next step—preventing blind retries, Domain Reload hangs, and broken scenes.

EES used 56% fewer tokens than Unity CLI.

Same 7 tasks. Both tools completed all 7.

Total tokens:

Unity CLI: 6.84 million

EES: 2.98 million

Token savings with EES:

61% fewer — Find GameObjects

40% fewer — Run EditMode tests

66% fewer — Add a C# component

30% fewer — Drag the mouse in Play Mode

62% fewer — Fix a bug

59% fewer — Save an animation keyframe

41% fewer — Capture the Inspector

https://github.com/creativeandcreative9/EditorEyeStandAlone/blob/main/benchmarks/2026-10-unity-cli-vs-ees/README.md

Editor Eye StandAlone gives your existing coding agent an observable, scriptable feedback loop inside the Unity Editor.

It can inspect the real project, run Editor C#, enter Play Mode, capture results, read what happened, and iterate. You spend less time acting as the agent's eyes and hands.

Inspect → change → run → observe → refine.

That is the missing loop between AI-generated code and a result you can actually check in Unity.

Delegate Unity Work, Not Just Script Writing

Diagnose against the live Editor

A prefab is missing. A material renders black. A button ignores clicks. A script throws a `NullReferenceException`.

Instead of diagnosing from your description alone, your agent can query the scene hierarchy, GameObjects, components, tags, layers, references, Console history, and compile errors. It can apply a fix, wait for compilation, inspect the new result, and continue if another change is needed.

You describe the symptom. The agent gathers the Unity evidence.

Check what the scene actually looks like

Compiling is not the same as working. A panel can be off-center, a character can clip through the floor, or a prop can appear behind the camera.

EES can capture the Scene View, Game View, or a named camera. A vision-capable coding agent can inspect the image, adjust the scene, and capture another result without asking you to take every screenshot.

Turn repetitive Editor work into one request

- Place props along a curve.

- Rename objects to match a convention.

- Swap a material across dozens of prefabs.

- Repair references after reorganizing folders.

- Toggle or configure a component across a scene.

Your agent can write and run C# directly inside the Editor, using the same public `UnityEditor`, `UnityEngine`, and `AssetDatabase` APIs available to Editor scripts. One general execution tool reaches far beyond a fixed list of convenience actions.

Run and verify, not just edit

Inspect — query hierarchies, objects, components, Console logs, and compile state.

Build — create and modify GameObjects, prefabs, materials, shaders, textures, ScriptableObjects, and other assets through live C#.

Play — enter, exit, pause, and step Play Mode.

Check — capture screenshots, assert runtime state, run NUnit tests, and read resulting values. Typed responses report applied and unchanged fields where relevant, so the agent has concrete state to work from.

Three Steps to Your First Editor Request

1. Import `EditorEyeStandalone_vX_X_X.unitypackage` and wait.

2. Once the import is complete, the **Editor Eye Standalone — Ready** window

appears in front of the Unity Editor. Enter the prompt shown in the window

into your AI coding agent.

3. Tell the AI what you want it to do. It will start working in Unity.

Setup writes the protocol to `EditorEye/AGENTS.md` and adds a short pointer to supported agent-instruction files. Existing files are appended to, not overwritten, and running setup again is safe.

EES works with filesystem-capable coding agents such as **Claude Code, Codex CLI, Cursor, and Gemini CLI**. You keep using the agent you already know.

No Server, No Open Port, No Executable

EES communicates through small JSON files inside your Unity project:

```text

<Project>/EditorEye/inbox/<request-id>.json    ← your agent writes a request

<Project>/EditorEye/outbox/<request-id>.json   ← Unity writes the response

```

There is no EES server to launch, network port to configure, executable to install, worker process to keep alive, or EES account to create.

Because requests live on disk, a request written while Unity is compiling or performing a Domain Reload waits for the Editor to become ready. There is no live socket connection for the reload to drop.

The package contains readable C# source, and its wire format is documented JSON. You can inspect, modify, or extend the system inside your own project.

Know the Fit Before You Buy

You need:

- Unity **2022.3 LTS or later**

Bring your own AI: EES does not include an AI model or make external AI API calls. Any subscription, API use, or data handling by your chosen coding agent is separate from EES.

Supported OS

Windows 11, macOS, and Linux. On non-Windows platforms, Game View capture renders only the camera output, because the OS window itself cannot be captured.

What You Get

- Approximately **27MB**

- One Package Manager dependency: `com.unity.nuget.newtonsoft-json`

- No EES subscription, license key, online activation, telemetry, or EES network traffic

- Readable Editor C# source and documented local JSON transport

- Verified with **0 compile errors** on a clean Unity 2022.3.22f1 and Unity 6000.3.7f1 project with the required dependency installed.

Give your coding agent a real Unity feedback loop.

For powerful AI-integrated software, choose EditorEye.