> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fetchserp.com/llms.txt
> Use this file to discover all available pages before exploring further.

# LLM Integration

> Super-charge your AI with real-time SEO data from fetchSERP

fetchSERP speaks **Model Context Protocol (MCP)**, making it plug-and-play with any LLM that supports tool calling.

## Supported Providers

<CardGroup cols={3}>
  <Card title="Claude API" icon="brain" href="#claude">
    Anthropic models with tool calling (v2025-05-14+)
  </Card>

  <Card title="OpenAI API" icon="globe-americas" href="#openai">
    GPT-4 / GPT-4o with `mcp` tool
  </Card>

  <Card title="Custom Agents" icon="puzzle-piece" href="#custom">
    Any framework that can call REST endpoints
  </Card>
</CardGroup>

## Claude API Example

```javascript title="claude.js" theme={null}
const claudeRequest = {
  model: 'claude-sonnet-4-20250514',
  messages: [
    { role: 'user', content: 'Give me keyword ideas for "seo audit"' }
  ],
  mcp_servers: [
    {
      type: 'url',
      url: 'https://www.fetchserp.com/mcp',
      name: 'fetchserp',
      authorization_token: process.env.FETCHSERP_API_TOKEN,
      tool_configuration: {
        enabled: true
      }
    }
  ]
};
```

## OpenAI API Example

```javascript title="openai.js" theme={null}
let response = await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [
    { role: 'user', content: 'Check my domain ranking for "project management tools"' }
  ],
  tools: [
    {
      type: 'mcp',
      server_label: 'fetchserp',
      server_url: process.env.MCP_SERVER_URL, // https://www.fetchserp.com/mcp
      headers: {
        Authorization: `Bearer ${process.env.FETCHSERP_API_TOKEN}`
      }
    }
  ]
});
```

## Custom Agent Flow

1. Call the `/mcp/schema` endpoint to fetch a JSON schema describing every fetchSERP tool.
2. Feed that schema to your LLM along with user instructions.
3. Execute tool calls returned by the LLM and stream the results back.

For an end-to-end reference implementation, check out the **fetchserp-mcp-server** on GitHub.

## Best Practices

* **One server per request** – List only the tools you actually need.
* **Stream responses** – Large SERP payloads benefit from HTTP chunked encoding.
* **Cache intelligently** – Many endpoints are cached for 15 minutes, so store UUIDs where applicable.

## Features Available via MCP

| Tool Name                       | REST Endpoint                          | What It Does                                                 |
| ------------------------------- | -------------------------------------- | ------------------------------------------------------------ |
| serp                            | `/api/v1/serp`                         | Structured SERP results from Google, Bing, Yahoo, DuckDuckGo |
| ranking                         | `/api/v1/ranking`                      | Domain ranking for a keyword                                 |
| serp\_html                      | `/api/v1/serp_html`                    | SERP results with full HTML content                          |
| serp\_text                      | `/api/v1/serp_text`                    | SERP results with extracted text                             |
| serp\_js                        | `/api/v1/serp_js`                      | JavaScript-rendered SERP with AI Overview                    |
| serp\_ai                        | `/api/v1/serp_ai`                      | AI Overview + AI Mode response                               |
| serp\_ai\_mode                  | `/api/v1/serp_ai_mode`                 | Cached US-only AI Mode                                       |
| keywords\_search\_volume        | `/api/v1/keywords_search_volume`       | Monthly search volume metrics                                |
| keywords\_suggestions           | `/api/v1/keywords_suggestions`         | Keyword ideas & metrics                                      |
| long\_tail\_keywords\_generator | `/api/v1/long_tail_keywords_generator` | Generate long-tail keywords                                  |
| page\_indexation                | `/api/v1/page_indexation`              | Check if a page is indexed                                   |
| backlinks                       | `/api/v1/backlinks`                    | Backlink data for a domain                                   |
| web\_page\_seo\_analysis        | `/api/v1/web_page_seo_analysis`        | Technical & on-page SEO audit                                |
| web\_page\_ai\_analysis         | `/api/v1/web_page_ai_analysis`         | AI-powered content analysis                                  |
| scrape                          | `/api/v1/scrape`                       | Raw HTML (no JS)                                             |
| scrape\_js                      | `/api/v1/scrape_js`                    | Custom JS extraction                                         |
| scrape\_js\_with\_proxy         | `/api/v1/scrape_js_with_proxy`         | JS extraction through proxy                                  |
| domain\_scraping                | `/api/v1/domain_scraping`              | Crawl an entire domain                                       |
| domain\_infos                   | `/api/v1/domain_infos`                 | DNS, WHOIS, SSL, tech stack                                  |
| domain\_emails                  | `/api/v1/domain_emails`                | Extract emails from SERPs                                    |
| moz                             | `/api/v1/moz`                          | Moz domain authority & metrics                               |

> These tools are auto-generated from our OpenAPI spec, so new endpoints will appear here automatically.

Need help? [Contact support](mailto:support@fetchserp.com) or jump in our Slack community.