Documentation

Head over to the install page and grab the .unitypackage file. Drop it into your Unity project and you're ready to be driven by an AI.

1. Double-click the downloaded .unitypackage (or use Assets → Import Package → Custom Package… from the Unity menu) to import it.
2. Once Unity finishes recompiling, setup is complete automatically. You'll see Window → Editor Eye in the menu bar — that means it worked.
3. Open Window → Editor Eye → Settings, paste the license key from your dashboard, and click Connect.

If you're using Claude Desktop, add Editor Eye to its config file as a tool Claude can call. Just paste the JSON below.

Config file location: %APPDATA%\Claude\claude_desktop_config.json (Windows) / ~/Library/Application Support/Claude/claude_desktop_config.json (Mac)

{
  "mcpServers": {
    "editor-eye": {
      "url": "https://api.editor-eye.app/mcp",
      "headers": {
        "Authorization": "Bearer your-license-key"
      }
    },
    "editor-eye-windows": {
      "command": "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe",
      "args": ["mcp-server"]
    }
  }
}

※ Replace {username} with your Windows username.
※ editor-eye-windows is Windows-only. Skip it on Mac / Linux.

If you live in the terminal with Claude Code, add the same block to the mcpServers section of ~/.claude/settings.json. You can drive Unity without ever leaving your editor.

{
  "mcpServers": {
    "editor-eye": {
      "url": "https://api.editor-eye.app/mcp",
      "headers": {
        "Authorization": "Bearer your-license-key"
      }
    },
    "editor-eye-windows": {
      "command": "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe",
      "args": ["mcp-server"]
    }
  }
}

Restart Claude Code and the editor-eye tools become available.

If your version of Cursor supports MCP, adding Editor Eye to the config may let you drive Unity from the Cursor chat. (We haven't verified this ourselves.)

1. Open .cursor/mcp.json in your project root (for per-project use) or ~/.cursor/mcp.json in your home directory (for every project).
2. Paste the same block shown for Claude Desktop in 1-2 — the mcpServers shape is identical.
3. Restart Cursor and the editor-eye tools become available.

{
  "mcpServers": {
    "editor-eye": {
      "url": "https://api.editor-eye.app/mcp",
      "headers": {
        "Authorization": "Bearer your-license-key"
      }
    },
    "editor-eye-windows": {
      "command": "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe",
      "args": ["mcp-server"]
    }
  }
}

※ Exact paths and enablement steps may change between Cursor versions — cross-check with the Cursor docs.

On VS Code 1.99 or later with the GitHub Copilot MCP feature enabled, the following config may let you call Editor Eye. (We haven't verified this ourselves.)

1. In VS Code settings.json, enable "github.copilot.chat.mcp.discovery.enabled": true.
2. Create .vscode/mcp.json in your project and paste the same content as 1-2.
3. Restart VS Code and Editor Eye tools should be callable from Copilot Chat.

{
  "mcpServers": {
    "editor-eye": {
      "url": "https://api.editor-eye.app/mcp",
      "headers": {
        "Authorization": "Bearer your-license-key"
      }
    },
    "editor-eye-windows": {
      "command": "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe",
      "args": ["mcp-server"]
    }
  }
}

※ MCP setting keys and enablement flags shift around as VS Code / Copilot evolve — cross-check with the VS Code docs.

If your Gemini CLI build supports MCP, the snippet below may get you connected. (We haven't verified this ourselves.)

1. Open ~/.gemini/settings.json (create it if missing).
2. Add the same mcpServers block as 1-2.
3. Restart Gemini CLI and the editor-eye tools should appear.

{
  "mcpServers": {
    "editor-eye": {
      "url": "https://api.editor-eye.app/mcp",
      "headers": {
        "Authorization": "Bearer your-license-key"
      }
    },
    "editor-eye-windows": {
      "command": "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe",
      "args": ["mcp-server"]
    }
  }
}

※ Paths and key names may change between Gemini CLI versions — check the Gemini CLI repo.

If the OpenAI Codex CLI supports MCP, this config may get you connected. (We haven't verified this ourselves.)

1. Add the MCP server definition to ~/.codex/config.toml (or pass it via the CODEX_MCP_SERVERS env var).
2. The shape mirrors the Claude Desktop mcpServers structure, just in TOML.
3. Restart Codex CLI and Editor Eye tools should be discovered.

# Append to the mcp_servers section of ~/.codex/config.toml
[mcp_servers.editor-eye]
url = "https://api.editor-eye.app/mcp"
headers = { Authorization = "Bearer your-license-key" }

[mcp_servers.editor-eye-windows]
command = "C:/Users/{username}/AppData/Local/EditorEye/editor-eye-windows-cli.exe"
args = ["mcp-server"]

Every instruction from the AI reaches Unity through one of two paths: editor-eye (cloud route) or editor-eye-windows (local route). Reading the scene, editing GameObjects, running scripts — those go through the cloud. Grabbing the screen, bringing windows forward — those stay local. The AI picks the right route for you.

AI (Claude Code)
  │
  ├─ editor-eye (all OS)
  │    └─ editor-eye-server (Railway cloud)
  │         └─ Unity Bridge (WebSocket)
  │              └─ Unity Editor
  │
  └─ editor-eye-windows (Windows only)
       └─ editor-eye-windows-cli.exe (local)
            └─ Unity Editor (direct IPC)

Editor Eye keeps the bridge between your AI and Unity alive on its own. Brief network blips, Unity restarts — none of those require you to do anything. Just three things to keep in mind.

gateway_core_screenshot grabs Unity's Game View or Scene View (the game itself, basically) and hands it to the AI. Since it works on Mac and Linux too, this is the one to try first.