Content
<div align="center">
<img src="./assets/serphouse-logo.png" alt="SERPHouse logo" width="80" />
# SERPHouse MCP Server
**Connect AI assistants to live SERP data, Google verticals, and SEO intelligence — powered by [SERPHouse](https://serphouse.com).**
Run Google, Bing, and Yahoo searches, resolve locations, and query Jobs, Local, Videos, and more — directly from Cursor, VS Code, Claude Desktop, or any MCP-compatible client. No custom API integration required.
<br />
[](https://www.typescriptlang.org/)
[](https://modelcontextprotocol.io/)
[](LICENSE)
[](https://insiders.vscode.dev/redirect/mcp/install?name=serphouse-mcp&config=%7B%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fmcp.serphouse.com%2Fmcp%22%2C%22headers%22%3A%7B%22SERPHOUSE_API%22%3A%22YOUR_SERPHouse_API_KEY%22%7D%7D)
[](cursor://anysphere.cursor-deeplink/mcp/install?name=serphouse-mcp&config=eyJ0eXBlIjoiaHR0cCIsInVybCI6Imh0dHBzOi8vbWNwLnNlcnBob3VzZS5jb20vbWNwIiwiaGVhZGVycyI6eyJTRVJQSE9VU0VfQVBJIjoiWU9VUl9TRVJQSG91c2VfQVBJX0tFWSJ9fQ==)
[](https://serphouse.com)
[](https://github.com/SERPHouse)
[](https://linkedin.com/company/serphouse)
</div>
---
## Table of Contents
- [Why SERPHouse MCP](#why-serphouse-mcp)
- [Quick Start (Hosted)](#quick-start-hosted)
- [What You Can Ask](#what-you-can-ask)
- [Tools Overview](#tools-overview)
- [Self-Host Locally](#self-host-locally)
- [Self-Host with Docker](#self-host-with-docker)
- [Use with Local Llama Models (Ollama)](#use-with-local-llama-models-ollama)
- [Troubleshooting](#troubleshooting)
- [Contributing](#contributing)
- [License](#license)
---
## Why SERPHouse MCP
| | |
|---|---|
| **21 MCP tools** | Live Google, Bing, and Yahoo SERP, Google verticals, and account lookups — all exposed with typed schemas |
| **Zero glue code** | Your assistant picks the right tool; you describe what you need in plain language |
| **Hosted or self-hosted** | Use the managed endpoint at `https://mcp.serphouse.com/mcp`, or run the server on your own infrastructure |
| **Built-in context** | MCP resources (`serphouse_capabilities`, `serphouse_constraints`, `serphouse_examples`) teach the AI usage rules automatically |
---
## Quick Start (Hosted)
The fastest path — no build step, no server to maintain.
**1.** Get your API key from the [SERPHouse Dashboard](https://www.serphouse.com).
**2.** Add the server to your MCP client config:
```json
{
"mcpServers": {
"serphouse": {
"url": "https://mcp.serphouse.com/mcp",
"headers": {
"SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
}
}
}
}
```
**3.** Start chatting. Ask your assistant to search Google, Bing, or Yahoo, look up locations, fetch jobs, or check your account — it will route to the correct tool.
> **One-click install:** Use the **Install in VS Code** or **Install in Cursor** badges above, then replace the placeholder API key with yours.
---
## What You Can Ask
For SEO teams, agencies, and SaaS marketers who need live search data inside their AI workflow — no dashboards, scripts, or context switching.
| Why you use it | Example prompt |
|----------------|----------------|
| **Track your rankings** | *"Where do we rank for 'crm software' on Google US desktop?"* |
| **Beat competitors** | *"Who owns the top 5 spots for 'project management software' in London?"* |
| **Own local search** | *"Top Google Local results for 'emergency plumber' in Chicago."* |
| **Discover keywords** | *"What does Autocomplete suggest for 'best saas for'?"* |
| **Monitor positions** | *"Run a mobile Bing SERP check for our brand in NYC and report our rank."* |
| **Multi-engine coverage** | *"Compare Yahoo and Google news results for 'electric vehicles'."* |
---
## Tools Overview
The server exposes **21 tools** across five categories. Google and Bing SERP requests require exactly one location field — `loc` (e.g. `Austin,Texas,United States`) or `loc_id` (from `serphouse_location_search`). Never send both or omit both on those endpoints. Yahoo SERP tools do not require location.
### Reference
| Tool | Description |
|------|-------------|
| `serphouse_domain_list` | Supported Google, Bing, and Yahoo search domains |
| `serphouse_language_list` | Language codes by search engine type |
| `serphouse_location_search` | Resolve city/country names to `loc_id` |
| `serphouse_account_info` | Account balance and usage |
### Google SERP
| Tool | Description |
|------|-------------|
| `serphouse_google_web` | Google web search |
| `serphouse_google_image` | Google image search |
| `serphouse_google_news` | Google news search |
| `serphouse_google_shop` | Google shopping search |
| `serphouse_serp_google_advanced` | Advanced Google SERP with extended parameters (up to 100 results) |
### Bing SERP
| Tool | Description |
|------|-------------|
| `serphouse_bing_web` | Bing web search |
| `serphouse_bing_image` | Bing image search |
| `serphouse_bing_news` | Bing news search |
### Yahoo SERP
| Tool | Description |
|------|-------------|
| `serphouse_yahoo_web` | Yahoo web search |
| `serphouse_yahoo_image` | Yahoo image search |
| `serphouse_yahoo_news` | Yahoo news search |
### Google Verticals
| Tool | Description |
|------|-------------|
| `serphouse_google_jobs` | Google Jobs search |
| `serphouse_google_autocomplete` | Google Autocomplete suggestions |
| `serphouse_google_videos` | Google Videos results |
| `serphouse_google_short_videos` | Google Short Videos (Shorts) |
| `serphouse_google_forums` | Google Forums results |
| `serphouse_google_local` | Google Local / Maps results |
---
## Self-Host Locally
Run the server on your machine for full control or local development.
```bash
git clone https://github.com/SERPHouse/serphouse-mcp.git
cd serphouse-mcp
npm install
npm run build
npm start
```
The server listens on `http://localhost:3000`. MCP endpoint: `POST /mcp`
Point your MCP client at the local instance:
```json
{
"mcpServers": {
"serphouse": {
"url": "http://localhost:3000/mcp",
"headers": {
"SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
}
}
}
}
```
### Commands
| Command | Description |
|---------|-------------|
| `npm run build` | Compile TypeScript to `dist/` |
| `npm start` | Run the HTTP server (`http://localhost:3000/mcp`) |
| `npm run start:stdio` | Run the stdio transport server (local MCP process mode) |
| `npm run dev` | Run the HTTP server with hot reload |
| `npm run dev:ins` | Launch MCP Inspector against the stdio server for debugging |
| `npm run typecheck` | Type-check without emitting files |
Health check: `GET http://localhost:3000/health`
### Server Options
| Variable | Default | Description |
|----------|---------|-------------|
| `PORT` | `3000` | HTTP listen port |
| `HOST` | `0.0.0.0` | HTTP bind address |
Authentication is header-only. Send `SERPHOUSE_API: <api_key>` on every `/mcp` request (configure it in your MCP client `headers`).
---
## Self-Host with Docker
Run the MCP server locally in Docker without installing Node.js.
```bash
docker compose up -d
```
The server runs at `http://localhost:3000/mcp`.
Connect your MCP client with the API key in headers:
```json
{
"mcpServers": {
"serphouse": {
"url": "http://localhost:3000/mcp",
"headers": {
"SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
}
}
}
}
```
---
## Use with Local Llama Models (Ollama)
Hook up your SERPHouse MCP Server to a local Llama model (e.g. Llama 3.1 or 3.2) using [`ollmcp`](https://github.com/jonigl/mcp-client-for-ollama) — an interactive terminal UI client that brings real-time search engine power straight to your local LLM workflow.
### Prerequisites
| Requirement | Notes |
|-------------|-------|
| **Tool-calling Llama model** | A model that supports tool calls (e.g. `llama3.1`, `llama3.2`) installed and running via [Ollama](https://ollama.com) |
| **Python 3.10+** | Required to run the `ollmcp` terminal client |
| **Node.js** | Required to run the SERPHouse MCP server via `npx` |
### 1. Install `ollmcp`
`ollmcp` is a Python client, so install it globally with `pip`:
```bash
pip install --upgrade ollmcp
```
### 2. Create a `config.json` file
The client needs a configuration profile that tells it how to launch the SERPHouse server and pass your API credentials. Create a `config.json` in your working folder:
```json
{
"mcpServers": {
"serphouse-mcp": {
"url": "https://mcp.serphouse.com/mcp",
"headers": {
"SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
}
}
}
}
```
> Replace `YOUR_SERPHouse_API_KEY` with your active key from the [SERPHouse Dashboard](https://www.serphouse.com).
### 3. Run the terminal client
Launch the interactive interface, passing your config profile and target model:
```bash
ollmcp --servers-json config.json --model llama3.1
```
###### Once the terminal UI loads:
1. Select your target Llama model from the model list.
2. Submit a prompt that needs live web data — e.g. *"Look up the top Google results for 'best developer tools' using SERPHouse."*
3. Watch your local Llama model call the SERPHouse tools and turn real-time SERP data into a conversational answer.
### Compatible Models
The following Ollama models work well with tool use:
- gemma4
- qwen3.5
- lfm2.5-thinking
- llama3.2
- mistral
For a complete list of Ollama models with tool use capabilities, visit the official [Ollama models](https://ollama.com/search?c=tools) page.
For models that can also process images returned by tools, see the Ollama [vision models](https://ollama.com/search?c=vision) page.
---
## Troubleshooting
| Issue | Fix |
|-------|-----|
| Missing API key | Send `SERPHOUSE_API` header on `/mcp` requests |
| Invalid key | Verify your key in the [SERPHouse dashboard](https://serphouse.com) |
| Credit exhausted | Check balance with `serphouse_account_info` |
| Location error | For Google and Bing tools, include `loc` or `loc_id` (not both); use `serphouse_location_search` to resolve IDs. Yahoo SERP tools do not require location. |
| Connection refused | Confirm the server is running and the URL/port is correct |
---
## Contributing
Contributions are welcome. Please keep changes focused and match existing code style.
```bash
git checkout -b feature/your-feature
npm install
# make changes
npm run typecheck
git commit -m "Add your feature"
git push origin feature/your-feature
```
Then open a Pull Request. Update this README if you change setup or configuration.
---
## License
[MIT License](LICENSE) — Copyright SERPHouse.
MCP Config
Below is the configuration for this MCP Server. You can copy it directly to Cursor or other MCP clients.
mcp.json
Connection Info
You Might Also Like
No recommendations available