SERPHouse MCP

---

Table of Contents

• Why SERPHouse MCP
• Quick Start (Hosted)
• Authentication
• What You Can Ask
• Tools Overview
• Self-Host Locally
• Self-Host with Docker
• Use with Local Llama Models (Ollama)
• Troubleshooting
• Contributing
• 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.

2. Add the server to your MCP client config using either header-based or URL-based authentication (see Authentication for details).

Header (recommended):

{
  "mcpServers": {
    "serphouse": {
      "url": "https://mcp.serphouse.com/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

URL path:

{
  "mcpServers": {
    "serphouse": {
      "url": "https://mcp.serphouse.com/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

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.

---

Authentication

The server accepts your SERPHouse API key in two ways. Use whichever fits your MCP client.

Method	Endpoint	How to pass the key
Header (recommended)	POST /mcp	SERPHOUSE_API: <api_key> request header
URL path	POST /{apiKey}/mcp	Embed the key in the URL path

Tools and resources can be discovered without a key. The API key is required only when calling SERPHouse API tools.

Option A: Header (recommended)

Best when your MCP client supports custom headers. Keeps the key out of URLs and server access logs.

{
  "mcpServers": {
    "serphouse": {
      "url": "https://mcp.serphouse.com/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

Self-hosted example:

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

Option B: URL path

Useful when your client only supports a URL and cannot send custom headers.

{
  "mcpServers": {
    "serphouse": {
      "url": "https://mcp.serphouse.com/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

Self-hosted example:

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

Note: URL-based auth places the API key in the request path, which may appear in proxy logs or browser history. Prefer the header method when possible.

---

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.

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 endpoints:

• POST /mcp — pass the API key via the SERPHOUSE_API header
• POST /{apiKey}/mcp — pass the API key in the URL path

Point your MCP client at the local instance (see Authentication for both options):

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

Or with URL-based auth:

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

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 supports both header and URL path. See Authentication.

---

Self-Host with Docker

Run the MCP server locally in Docker without installing Node.js.

docker compose up -d

The server runs at http://localhost:3000/mcp (header auth) or http://localhost:3000/{apiKey}/mcp (URL auth).

Connect your MCP client using either method from Authentication:

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

Or:

{
  "mcpServers": {
    "serphouse": {
      "url": "http://localhost:3000/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

---

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 — 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
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:

pip install --upgrade ollmcp

2. Create a config.json file

The client needs a configuration profile that tells it how to connect to the SERPHouse server and pass your API credentials. Create a config.json in your working folder using either auth method from Authentication:

{
  "mcpServers": {
    "serphouse-mcp": {
      "url": "https://mcp.serphouse.com/mcp",
      "headers": {
        "SERPHOUSE_API": "YOUR_SERPHouse_API_KEY"
      }
    }
  }
}

Or with URL-based auth:

{
  "mcpServers": {
    "serphouse-mcp": {
      "url": "https://mcp.serphouse.com/YOUR_SERPHouse_API_KEY/mcp"
    }
  }
}

Replace YOUR_SERPHouse_API_KEY with your active key from the SERPHouse Dashboard.

3. Run the terminal client

Launch the interactive interface, passing your config profile and target model:

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 page.

For models that can also process images returned by tools, see the Ollama vision models page.

---

Troubleshooting

Issue	Fix
Missing API key	Send SERPHOUSE_API header on POST /mcp, or use POST /{apiKey}/mcp with the key in the URL path
Invalid key	Verify your key in the SERPHouse dashboard
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.

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 — Copyright SERPHouse.