Google Trends MCP

A local MCP server that lets Claude (or any MCP client) query Google Trends directly — search interest over time, related queries and topics, regional breakdowns, and real-time trending searches — so you can do market research inside a conversation instead of tab-switching to trends.google.com and pasting screenshots back in.

Built for personal/self-use market research. No API key required — Google Trends has no official public API, so this wraps pytrends, the standard unofficial Python client, in an MCP server.

Tools

All tools default geo="IN" (India) unless noted — pass geo="" for worldwide, or any ISO country code ("US", "GB", etc.). timeframe accepts pytrends' format, e.g. "today 12-m", "today 5-y", "now 7-d", or an explicit range "2024-01-01 2024-06-01".

interest_over_time(keywords, timeframe="today 12-m", geo="IN")

Relative search interest (0–100) over time for up to 5 keywords, compared side by side. Keywords beyond the first 5 are silently dropped. Each record includes isPartial — true on the most recent data point means that period isn't finished yet and its value is provisional; don't read a dip on that point as a real trend change.

related_queries(keyword, timeframe="today 12-m", geo="IN")

Top and rising related search queries for a single keyword. Returns {"top": [...], "rising": [...]}, each a list of {"query": ..., "value": ...} records. top values are 0–100 relative interest. rising values are percent increase — except a value of 5000%, which is Google's "Breakout" marker for explosive growth from a near-zero baseline, not a literal percentage.

related_topics(keyword, timeframe="today 12-m", geo="IN")

Same as related_queries, but topic clusters (Google's own topic groupings) instead of raw query strings — records have topic_title and topic_type alongside value. Same Breakout convention applies to rising.

interest_by_region(keyword, timeframe="today 12-m", geo="IN")

Search interest for a keyword broken down by state/region within the given geo. Returns a list of {"geoName": ..., "<keyword>": 0-100} records, one per region.

trending_now(geo="india")

Today's top trending searches for a country. Note the geo format is different here — it's a full lowercase country name ("india", "united_states"), not an ISO code like the other four tools. This is a real inconsistency in Google's own endpoints, not a bug.

Known limitation: as of this writing, trending_now fails with an HTTP 404. Google appears to have retired the legacy endpoint (hottrends/dailytrends/realtimetrends) that pytrends' trending-search methods depend on — confirmed by testing all three variants pytrends offers. This is an upstream issue, not fixable in this codebase; it fails cleanly with a readable error string rather than crashing. The other 4 tools use a different, still-functional endpoint family and are unaffected. If Google restores the endpoint or pytrends patches around it, this will start working again with no changes needed here.

All tools catch failures (rate limits, network errors, the above) and return a plain error string instead of crashing — Google Trends is a scraped endpoint, not a stable API, so this is expected behavior, not exceptional.

Setup

Requires Python 3.10+ and uv.

git clone https://github.com/jain-eshan/google-trends-mcp.git
cd google-trends-mcp
uv sync

Register with Claude Code

claude mcp add google-trends -- uv run --directory /absolute/path/to/google-trends-mcp server.py

Verify it connected:

claude mcp list

You should see google-trends listed as ✔ Connected. Start a new Claude Code conversation after registering — sessions already running won't pick up a newly added server.

Usage

Once registered, just ask Claude to use it — e.g.:

"Use the google-trends MCP to compare interest in 'lab grown diamonds' vs 'diamond jewellery' in India over the last 12 months, and show me related queries."

Optional: /trends skill

This repo includes a Claude Code skill at .claude/skills/trends/SKILL.md that wraps the raw tools into a research-and-synthesize workflow — it decides which tools are relevant to your topic and writes up a plain-language summary instead of dumping raw JSON. If you're using Claude Code, this skill is picked up automatically from this repo; just run:

/trends <your topic>

Design notes

• Data only, no synthesis in the server. Every tool returns raw, structured data — the interpretation (is this trend real, what does a Breakout marker mean here, what's worth flagging) happens in the calling conversation, not baked into the server. This keeps the server simple and lets whatever's calling it (Claude, another MCP client) apply its own judgment.

• No dependencies beyond mcp[cli] and pytrends. No database, no config file, no API key.

• No formal test suite. This wraps a scraped third-party endpoint; a test suite would mostly be testing pytrends and Google's current response shape, not this code. Each tool was verified against live Google Trends data during development instead.

License

MIT — see LICENSE.