Google Trends MCP

What is Google Trends MCP?

Google Trends MCP is a Model Context Protocol server that gives AI assistants structured Google Trends data. It supports batch keyword research (up to 50 keywords per batch), cross-batch normalization for comparability, and production-level SaaS deployment. Built with FastMCP, it works with Claude Desktop, ChatGPT, Cursor, Perplexity, and other major AI clients.

How to use Google Trends MCP?

Install Python 3.11+, clone the repo, create a virtual environment, install dependencies, then run python server.py for local stdio mode. For a web service, use --transport streamable-http with host and port. Deploy publicly with Docker Compose. Configure client MCP entries with the endpoint URL and an API key (obtained from the dashboard after Google OAuth login).

Key features of Google Trends MCP

• Four MCP tools: compare_keywords, research_keywords, research_keywords_async, get_research_task
• Cross-batch normalization using an anchor keyword for comparability
• Supports up to 50 keywords in a single research request
• Multiple access methods: MCP protocol, REST API, and GPT Actions
• Production-grade frontend with bilingual SaaS landing page, dashboard, and documentation
• Docker Compose deployment with automatic HTTPS and Google OAuth authentication

Use cases of Google Trends MCP

• Compare trending keywords over time to inform content strategy
• Batch research dozens of keywords for competitive analysis and SEO planning
• Integrate real-time Google Trends data into AI-driven market research workflows

FAQ from Google Trends MCP

What should I do if I hit a Google Trends 429 rate limit?

The server has built-in 1–3 second random delays and backoff retries. If you still hit limits, reduce call frequency, set an HTTPS proxy, or switch to the trendspy provider.

Why can’t I use print() in stdio mode?

In stdio mode, stdout carries MCP JSON-RPC protocol messages; ordinary text would break the protocol. Logs must be written to stderr. Only --smoke-test mode allows JSON output to stdout.

How is compare_keywords different from research_keywords?

compare_keywords returns raw Google Trends batch-relative scores for up to 5 keywords. research_keywords returns cross-batch normalized estimates using an anchor keyword, suitable for directional research across many keywords.

How fast is a batch research request?

With 2–4 concurrent threads and random delays, 10 keywords take about 5–10 seconds; 50 keywords take about 15–30 seconds. In-process caching (default 15 minutes) skips network calls for repeated requests.

What if I don’t have Google OAuth for local development?

When GOOGLE_CLIENT_ID is not set, the server enters dev mode. Visit /auth/dev-login to log in with a virtual account — no Google account needed.