AG
Gemini & Antigravity v2.0 Agentic Engineering Guide
Gemini 3 Models IDE Integration Verified: 2026-10

VS Code & Cursor Deep Integration with Gemini 3: Seamless AI Pair Programming

Configure Google Gemini 3 models in Visual Studio Code and Cursor IDE: From Roo Code and Continue to custom API endpoints, context rules, and .geminiignore.

Author: Google GenAI Evangelist Team • Google GenAI & Antigravity 2.0 Engineering Spec

Cursor & VS Code Gemini 3 Configuration

.continue/config.json

{
  "models": [
    {
      "title": "Gemini 3 Pro (DeepMind)",
      "provider": "gemini",
      "model": "gemini-3-pro-preview",
      "apiKey": "${GEMINI_API_KEY}"
    },
    {
      "title": "Gemini 3 Flash (Fast)",
      "provider": "gemini",
      "model": "gemini-3-flash-preview",
      "apiKey": "${GEMINI_API_KEY}"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Gemini 3 Flash Autocomplete",
    "provider": "gemini",
    "model": "gemini-3-flash-preview",
    "apiKey": "${GEMINI_API_KEY}"
  }
}

Modern IDEs Powered by Gemini 3

With Gemini 3’s fast Time-To-First-Token (TTFT) and massive 2M token context capacity, pairing it with your favorite development environment provides instant tab autocompletion and high-accuracy repository-wide refactoring.

Whether you rely on classic Visual Studio Code or the AI-native Cursor IDE, setup takes less than two minutes.


Method 1: VS Code Integration via Continue

Continue is the premier open-source AI extension for VS Code:

  1. Search for and install Continue - Codestral, Claude, and more from the VS Code Marketplace.
  2. Open ~/.continue/config.json.
  3. Copy the artifact configuration above and supply your GEMINI_API_KEY.
  4. Save and trigger Gemini 3 in your sidebar via Cmd + L (macOS) or Ctrl + L (Windows/Linux).

Method 2: Cursor IDE Integration

Cursor supports custom Gemini API endpoints natively:

  1. Open Cursor Settings (Cmd + , or top-right gear icon) and select Models.
  2. Locate the Google API Key field and paste your key from Google AI Studio.
  3. Under Model Names, click + Add Model and input the official identifiers:
    • gemini-3-pro-preview
    • gemini-3-flash-preview
  4. Select gemini-3-pro-preview in your Cursor Chat dropdown to begin code generation.

Best Practice: Context Exclusion with .geminiignore

Prevent large build artifacts, packages, and logs from clogging your context window by creating a .geminiignore in your repository root:

node_modules/
dist/
.astro/
*.log
.git/
package-lock.json

Gemini-powered extensions will respect this file, ensuring fast and accurate completions.