Review Guru MCP Server

An MCP server that exposes Sri Lankan business reviews and doctor profiles from Review Guru to Claude Desktop, Cursor, Cline, or any MCP-aware AI agent.

No API key. No scraping. Calls a public, cached, rate-limited HTTPS endpoint.

What you get

Tool	What it does
list_businesses(city?, category?, min_rating?, sort?, limit?)	Ranked list of businesses. Sort: top, most-reviewed, newest.
get_business(slug)	Full profile: address, phone, hours, categories, rating breakdown, top 10 reviews.
get_reviews(slug, sort?, limit?)	Paginated reviews. Sort: newest, highest, lowest, helpful.
search(query, limit?)	Full-text search (FTS5) across every business and doctor.
list_categories()	Top-level categories + sub-categories with slugs.
list_cities()	Sri Lankan cities with business counts.

Plus a reviewguru://about resource that returns live counts.

Data covers restaurants, hotels, shops, salons, hospitals, Sri Lankan doctors with specialty + practice locations, and more — all with patient/ customer reviews.

Install

With uv (recommended)

uvx reviewguru-mcp

With pipx

pipx install reviewguru-mcp
reviewguru-mcp

With pip

pip install reviewguru-mcp
reviewguru-mcp

The server speaks MCP over stdio — invoke it from any MCP client config.

Wire it up

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "reviewguru": {
      "command": "uvx",
      "args": ["reviewguru-mcp"]
    }
  }
}

Restart Claude Desktop. The 🔌 icon in chat will show the Review Guru tools.

Cursor

Settings → MCP → Add new MCP server:

• Command: uvx
• Args: reviewguru-mcp

Cline (VS Code)

Cline → Settings → MCP Servers → Add:

{
  "reviewguru": {
    "command": "uvx",
    "args": ["reviewguru-mcp"]
  }
}

Try it

Once wired up, ask your assistant things like:

Find me three highly-rated cardiologists in Colombo and tell me which hospital each one consults at.

What are people saying about Ministry of Crab? Pull the most recent five reviews.

I'm a vegetarian visiting Galle for a weekend — recommend two restaurants.

The model will pick the right tools and stitch the answers together.

Configuration

Environment variables (all optional):

Variable	Default	Effect
REVIEWGURU_API	https://reviewguru.lk	API base URL — point at staging or a fork.
REVIEWGURU_URL	https://reviewguru.lk	Used to render business URLs in tool output.
DATABASE_URL	(none)	If set to an existing SQLite path, server uses local DB instead of HTTP. Maintainer-only.

Rate limits + acceptable use

The public API is shared and lightly rate-limited (60 req/min per IP). For heavier usage, set REVIEWGURU_API to your own deployment.

Reviews are licensed for citation with attribution to Review Guru and a link back to the specific business URL. See reviewguru.lk/llms.txt for full LLM-usage guidelines.

Self-hosting

Want to fork? The server is one Python file (server.py). It auto-detects:

• HTTP mode (default) — calls /api/v1 endpoints. Works anywhere.
• SQLite mode — if data/reviewguru.db is present in the parent dir. Used by the maintainers for sub-millisecond local queries.

git clone https://github.com/kapruka/reviewguru-mcp
cd reviewguru-mcp
pip install -e .
reviewguru-mcp

Links

• 🌐 Site — https://reviewguru.lk
• 🔌 Public API — https://reviewguru.lk/api/v1
• 📜 OpenAPI spec — https://reviewguru.lk/api/openapi.json
• 📄 LLM usage policy — https://reviewguru.lk/llms.txt
• 🐛 Issues — https://github.com/kapruka/reviewguru-mcp/issues

License

MIT — see LICENSE.