Google Trends MCP
Google Trends MCP
Un servidor MCP local que permite a Claude (o a cualquier cliente MCP) consultar Google Trends directamente — interés de búsqueda a lo largo del tiempo, consultas y temas relacionados, desgloses por regiones y búsquedas de tendencia en tiempo real — para que puedas hacer investigación de mercado dentro de una conversación, sin cambiar de pestaña a trends.google.com y tener que pegar capturas de pantalla.
Está pensado para investigación de mercado de uso personal. No requiere clave de API — Google Trends no tiene una API pública oficial, así que esto envuelve pytrends, el cliente estándar no oficial de Python, en un servidor MCP.
Herramientas
Todas las herramientas usan geo="IN" (India) por defecto, salvo que se indique lo contrario — pasa geo="" para todo el mundo, o cualquier código de país ISO ("US", "GB", etc.). timeframe acepta el formato de pytrends, p. ej. "today 12-m", "today 5-y", "now 7-d", o un rango explícito "2024-01-01 2024-06-01".
interest_over_time(keywords, timeframe="today 12-m", geo="IN")
Interés de búsqueda relativo (0–100) a lo largo del tiempo para hasta 5 palabras clave, comparadas una al lado de la otra. Las palabras clave que superen las 5 primeras se descartan silenciosamente. Cada registro incluye isPartial — un valor true en el punto de datos más reciente significa que ese período aún no ha terminado y su valor es provisional; no interpretes una caída en ese punto como un cambio real de tendencia.
related_queries(keyword, timeframe="today 12-m", geo="IN")
Consultas de búsqueda relacionadas más populares y en aumento para una sola palabra clave. Devuelve {"top": [...], "rising": [...]}, cada uno una lista de registros {"query": ..., "value": ...}. Los valores de top son interés relativo de 0 a 100. Los valores de rising son el incremento porcentual — excepto un valor de 5000%, que es el marcador "Breakout" de Google para indicar crecimiento explosivo desde una base casi nula, no un porcentaje literal.
related_topics(keyword, timeframe="today 12-m", geo="IN")
Lo mismo que related_queries, pero con agrupaciones de temas (las propias agrupaciones de Google) en lugar de cadenas de consulta en bruto — los registros incluyen topic_title y topic_type junto a value. La misma convención de Breakout se aplica a rising.
interest_by_region(keyword, timeframe="today 12-m", geo="IN")
Interés de búsqueda de una palabra clave desglosado por estado/región dentro del geo indicado. Devuelve una lista de registros {"geoName": ..., "<keyword>": 0-100}, uno por cada región.
trending_now(geo="india")
Las búsquedas con más tendencia de hoy para un país. Nota: el formato de geo aquí es diferente — es el nombre completo del país en minúsculas ("india", "united_states"), no un código ISO como en las otras cuatro herramientas. Es una inconsistencia real en los propios endpoints de Google, no un error.
Limitación conocida: en el momento de crear escribir esto,
trending_nowdevuelve un HTTP 404. Google parece haber retirado el endpoint heredado (hottrends/dailytrends/realtimetrends) del que dependen los métodos de búsqueda de tendencias de pytrends, confirmado al probar las tres variantes que ofrece pytrends. Es un problema upstream que no puede solucionarse en este código; falla limpiamente con una cadena de error legible en lugar de colgarse. Las otras 4 herramientas utilizan una familia distinta de endpoints, que aún funciona y no se ve afectada. Si Google restaura el endpoint o pytrends lo soluciona con un parche, esto volverá a funcionar sin necesidad de cambios aquí.
Todas las herramientas capturan los fallos (límites de petición, errores de red, el caso anterior) y devuelven una cadena de error simple en lugar de bloquearse — pot hub años since Google Trends no es un endpoint estable, sino una API raspada; este es un comportamiento esperado, no algo excepcional.
Related MCP server: Google Trends MCP Server
Configuración
Requiere Python 3.10+ y uv.
git clone https://github.com/jain-eshan/google-trends-mcp.git
cd google-trends-mcp
uv syncRegistro con Claude Code
claude mcp add google-trends -- uv run --directory /absolute/path/to/google-trends-mcp server.pyVerifica que se ha conectado:
claude mcp listDeberías ver google-trends en la lista como ✔ Connected. Inicia una conversación nueva de Claude Code después del registro: las sesiones ya abiertas no lo detectarán un servidor recién agregado.
Uso
Una vez registrado, solo pide a Claude usarlo — por ejemplo:
"Usa el de google-trends MCP para comparar el interés en 'labor grown diamonds' vs 'diamond jewellery' in India durante los últimos 12 meses, y muéstrame las consultas relacionadas."
Opcional: skill /trends
Esta repo incluye una skill de Claude Code en .claude/skills/trends/SKILL.md que envuelve las herramientas en bruto en un flujo de trabajo de investigación y síntesis — decide qué herramientas son relevantes para tu tema y escribe un resumen natural, en vez de volcar el JSON en bruto. Si usas Claude Code, la skill se detecta automáticamente desde este repo; solo ejecuta:
/trends <your topic>Notas de diseño
Solo datos, sin síntesis en el servidor. Cada herramienta devuelve datos crudos y estructurados; a interpretación (si la tendencia es real, qué significa un marcador Breakout, qué se merece señalar) ocurre en la conversación que llama, no está incrustada en las piezas de escrita. Esto simplifica el servidor y deja que el que estén llamando (Claude, otro cliente MCP) aplique su buen juicio.
Sin dependencias específicas aparte de
mcp[cli]ypytrends. No se usa base de datos, ni archivo de config, ni un clave.No hay suite de pruebas formal. Esto envuelve un endpoint de terceros procesado con scripting; una suite de pruebas probaría sobre todo estilo de pytrends y la forma actual de respuesta de Google, no este código. En su lugar, cada herramienta se verificó con datos reales Google Trends durante el desarrollo.
Licencia
MIT — consulta LICENSE.
Available Tools
11 toolsapp_store_appsA
Existing iPhone/iPad apps matching a query, from Apple's public iTunes Search API - who already serves this need, and how many people rate them (rating count is a rough proxy for user base). No setup needed.
Args: query: search terms, e.g. "meal planner" or "invoice". country: ISO country code for the store, e.g. "US" (default), "IN", "GB". limit: max apps, capped at 50.
Returns: A list in App Store relevance order, each with "name", "developer", "rating", "rating_count", "price", "genre", "released", "last_updated", "url".
Inputs: query (required), limit, country.
builder_activityA
Are people actually building and shipping in this space? Hacker News stories
(launches, Show HNs, debates) plus GitHub repositories - a builder signal, as opposed
to search interest or community chatter. Works with no setup; a GITHUB_TOKEN saved via
/gutcheck setup raises GitHub's limit from 10 to 30 searches a minute.
For Product Hunt launches, use web search restricted to producthunt.com instead - Product Hunt's API has no search endpoint.
Args: query: search terms, e.g. a product category ("habit tracker") or problem.
Returns: {"hn": {"total": int, "stories": [...]}, "github": {"total": int, "repos": [...]}}. Each story: title, points, num_comments, url, created_at. Each repo (top 10 by stars): name, description, stars, last_push, url.
Inputs: query (required).
company_registrationA
Company registration lookup via OpenCorporates - registration facts only (incorporation date, status, company number). Does NOT cover funding, valuation, or traction data - no free API exists for that (see README for why).
Requires a free OpenCorporates API token: as of 2026 OpenCorporates requires a
token on every request, even on the free tier (roughly 50 requests/day, 200/month).
Register at https://opencorporates.com/api_accounts/new and save it as
OPENCORPORATES_API_TOKEN via /gutcheck setup.
Args: name: company name to search for. jurisdiction: optional OpenCorporates jurisdiction code (e.g. "in", "us_de") to narrow results.
Returns: A list of up to 5 matches, each containing "company_name", "jurisdiction_code", "incorporation_date", "company_number", "current_status", "opencorporates_url". Empty list if no matches. A setup-instructions string if OPENCORPORATES_API_TOKEN is unset, or OpenCorporates' own rejection message if the token is invalid/expired.
Inputs: name (required), jurisdiction.
interest_by_regionA
Search interest for a keyword broken down by state/region within a country, or by country when geo is "" (worldwide).
Args: keyword: a single search term. timeframe: pytrends timeframe string, e.g. "today 12-m". geo: ISO country code (e.g. "US", "IN"), or "" for worldwide (default). response_format: "concise" (default) returns only the top 10 regions by interest, rounded to whole numbers. "full" returns every region, unrounded.
Returns: A list of records, one per state/region within the specified geo, each containing: - "geoName": the state/region name (e.g. "California"), or the country name when worldwide - A column with the keyword name as the key: relative search interest (0-100 scale) for that region. Higher values indicate higher relative interest in that region compared to others in the same country. This is Google Trends' standard region-relative scale.
Inputs: keyword (required), geo, timeframe, response_format.
interest_over_timeA
Relative Google search interest (0-100) over time for up to 5 keywords, compared side by side.
Args: keywords: 1-5 search terms to compare. Only the first 5 are used; additional keywords are silently dropped. timeframe: pytrends timeframe string, e.g. "today 12-m", "today 5-y", "now 7-d", or "YYYY-MM-DD YYYY-MM-DD". geo: ISO country code (e.g. "US", "IN", "GB"), or "" for worldwide (default). response_format: "concise" (default) covers the whole requested timeframe, thinned to about 26 evenly spaced points (first and last always kept) and rounded to whole numbers, so the shape of the window is intact at a low token cost. "full" returns every point unrounded - use it when you need week-by-week detail.
Returns: A list of records, one per date, each containing: - "date": ISO date string - "isPartial": present and true only when the period is incomplete (most recent point) - omitted when false in "concise" mode, since false is the common case. - One numeric key per keyword (0-100 relative interest value)
Inputs: keywords (required), geo, timeframe, response_format.
news_coverageA
Recent news coverage from Google News - is the press writing about this, and what angle are they taking? Useful for spotting funding rounds, regulation, launches, and whether a topic is getting more or less attention. No setup needed.
Args:
query: search terms.
geo: ISO country code for the news edition, e.g. "US" (default), "IN", "GB".
recent: how far back - "7d", "1m", "1y" (default). Uses Google News' own when: filter.
limit: max headlines to return (Google returns up to ~100; "total" counts all of them).
Returns: {"total": int, "last_30_days": int, "headlines": [{"title", "source", "published", "url"}]}, newest first. "total" and "last_30_days" are counts within Google's capped result set, so treat them as a rough attention gauge, not an exact article count.
Inputs: query (required), geo, limit, recent.
reddit_signalA
What real people on Reddit are saying, asking, or complaining about - qualitative community signal, as opposed to Trends/Wikipedia's passive search/reading numbers.
Works with no setup via Reddit's public search feed (which returns post text but not
scores, and rate-limits after a burst of requests). If REDDIT_CLIENT_ID and
REDDIT_CLIENT_SECRET are saved via /gutcheck setup, uses the official API instead,
which adds score and comment counts.
Args: query: search terms. subreddits: optional subreddit names to restrict the search to (e.g. ["startups", "SaaS"]). limit: max results, capped at 100. time_filter: "week", "month", "year" (default), or "all".
Keep the query short (2-4 words) and pass subreddits when obvious communities exist.
Reddit's keyless search matches loosely on long questions, so results that share no
meaningful word with the query are dropped and the rest are ranked by how many query
words they match.
Returns: A list of records with "title", "subreddit", "snippet" (first ~300 chars of the post body), "url", "created". "score" and "num_comments" are included only in API mode.
Inputs: query (required), limit, subreddits, time_filter.
wikipedia_pageviewsA
Monthly Wikipedia pageview counts for an article - a free, no-auth reference/reading interest signal that complements Google Trends' search-interest signal. The two diverging (e.g. a term trending in search but flat on Wikipedia) can itself be a signal worth flagging.
Args: article: an English Wikipedia article title, e.g. "Artificial_intelligence" or "Machine learning" (spaces are handled automatically). timeframe: how far back to request, as a simple duration - "P1Y" (1 year, default), "P6M" (6 months), "P90D" (90 days). Only whole Y/M/D forms are supported. response_format: "concise" (default) returns only the most recent 12 months. "full" returns the entire requested timeframe.
Returns: A list of records, one per month, each containing: - "month": "YYYY-MM" - "views": total pageviews that month (all access methods, human traffic only - bot traffic is excluded by Wikimedia's "user" agent filter)
Inputs: article (required), timeframe, response_format.
youtube_videosA
YouTube videos matching a query, with view and comment counts - shows how much people watch content about a topic (tutorials, reviews, "I tried X" videos), which is often a stronger consumer-interest signal than search volume.
Requires a free YOUTUBE_API_KEY (Google Cloud, YouTube Data API v3), saved via
/gutcheck setup. Each call uses about 101 of the free 10,000 daily quota units.
Args: query: search terms. limit: max videos, capped at 25. published_after_days: only videos from the last N days; omit for any time.
Returns: A list in YouTube relevance order, each with "title", "channel", "published", "views", "likes", "comments", "url". A setup-instructions string if no key is saved.
Inputs: query (required), limit, published_after_days. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 11 tools
Each tool targets a distinct data source or dimension: Google Trends topics, queries, time series, regional breakdown, plus Wikipedia, OpenCorporates, Reddit, HN/GitHub, Google News, App Store, and YouTube. Even within Trends, related_topics vs related_queries are clearly separated by content type. No ambiguity.
All tool names follow a consistent snake_case noun phrase pattern (e.g., interest_over_time, reddit_signal, builder_activity). No mixing of conventions or verb styles, making the set predictable and easy to navigate.
With 11 tools covering a wide but focused range of market research signals, the count is well-scoped. Each tool contributes a unique data source, and none feel redundant or missing for the server's stated purpose.
The surface covers search interest, related topics/queries, regional data, reference interest (Wikipedia), community chatter (Reddit), builder activity (HN/GitHub), news coverage, app ecosystem, and YouTube viewership. Minor gaps exist (e.g., no global 'trending now' without a keyword, no multi-keyword regional comparison), but core workflows are well supported.
Maintenance
Related MCP Connectors
Google Trends: Search, Images, News, Shopping over time, growth metrics. Free key at trendsmcp.ai
Google Trends for agents: interest over time, by region, rising queries, Trending Now. No API key.
Google Trends realtime & entities
Bulk Google Trends: interest over time, regions, rising queries and plain-English trend summaries.
Related MCP Servers
- AlicenseBqualityDmaintenanceProvides access to Google Trends data including status, trending questions, and trending topics via MCP tools.2335 npm36MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to access Google Trends data for comparing keywords, discovering trending searches, and analyzing regional interest through natural language.2MIT
- AlicenseAqualityDmaintenanceProvides free Google Trends data (interest over time, term comparison, related queries, trending now, regional breakdown) to MCP-compatible AI clients without needing an API key.5134 npm2MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to query Google Trends data, including search interest over time, regional breakdowns, trending searches, and keyword ideas.MIT