Genkit MCP

---
title: Search Grounding and URL Context
description: read this to understand how to ground results on Google search or automatically ingest URLs in prompts to understand their contents
---

When using Gemini models, you can add configuration to automatically ground the prompt in Google search:

```ts
const {text} = await ai.generate({
  model: googleAI.model('gemini-2.5-flash'),
  // this includes the "built-in" search grounding tool
  config: {
    tools: [
      {googleSearch: {}},
    ]
  }
})
```

If your prompt contains URLs that you want the model to read and understand, you can add the URL context tool:

```ts
const {text} = await ai.generate({
  model: googleAI.model('gemini-2.5-flash'),
  // this includes the "built-in" url fetching tool
  config: {
    tools: [
      {urlContext: {}},
    ]
  }
})
```

You can use either or both of these tools to improve the "groundedness" of your generated model responses.