Data scientists working with consumer behavior, market signals, or social media research need clean, normalized, multi-source trend data - not a browser extension or a fragile scraper. Trends MCP delivers structured time series from Google, TikTok, Reddit, Amazon, Wikipedia, and 10 more sources via MCP or HTTP, ready for notebooks, pipelines, and AI-agent workflows.
Most trend data tooling is built for marketers and content teams, not analysts. Google Trends has a manual web interface. pytrends provides a Python wrapper but returns relative-only data and breaks on Google frontend changes. Social media platforms either have no API or rate-limit aggressively. The result is that data scientists who need clean, normalized, multi-source trend data for modeling spend a disproportionate amount of time on data collection infrastructure.
Trends MCP addresses this from a different angle. It is a managed data service that handles collection, normalization, and delivery - the analyst gets structured JSON time series covering 15+ sources, ready for notebooks or pipelines.
The get_trends endpoint returns weekly or daily time series for any keyword on any source. Each data point includes:
date - ISO 8601 timestampnormalized_value - 0-100 scale, consistent across all platformsabsolute_volume_estimate - calibrated raw volume where availabledata_quality_score - coverage and reliability indicator (0-1)The data_quality_score is particularly useful for modeling. It flags data points where coverage is sparse or where the normalization confidence is lower - allowing systematic filtering rather than manual inspection. A score below 0.5 on a given data point is worth excluding from regression inputs.
get_growth computes period-over-period growth for preset windows (7D, 14D, 1M, 3M, 6M, 1Y, YTD) or custom date ranges. The response includes exact start/end dates, start/end volumes, percentage change, and direction - structured to be used directly as features without further calculation.
Raw trend data from different platforms is not directly comparable. A TikTok hashtag with 500,000 uses and a Google search term with 500,000 monthly searches represent very different audience behaviors and very different scales. Comparing them directly as features in a model produces misleading results.
Trends MCP normalizes all sources to a consistent 0-100 scale calibrated to each platform's native volume distribution. This makes cross-platform features genuinely comparable - a normalized value of 60 on TikTok and 60 on Google reflect proportionally equivalent levels of interest on their respective platforms. For models where the relative momentum across platforms is the signal (e.g., is this trend leading on TikTok before Google?), normalized comparability is a prerequisite.
See the dedicated page on normalization methodology for the full technical approach.
Trends MCP is MCP-native, but the underlying data is accessible via standard HTTP for Python pipelines that do not use an AI agent:
import requests
headers = {"Authorization": "Bearer YOUR_API_KEY"}
The JSON response is clean and consistent - no HTML artifacts, no schema changes without versioning. Error payloads are structured and include the parameter that caused the issue, which reduces debugging time in automated pipelines.
A persistent problem with trend data research is reproducibility. Google Trends normalizes data relative to the query time window and the set of comparison terms, which means the same query run on different dates returns different values. Screenshots and manual exports have no documented provenance.
Trends MCP's API-based access returns consistent historical data for the same query parameters. The timestamped JSON responses can be stored alongside analysis code, making the data collection step auditable. For research that may be reviewed or replicated - academic papers, investment memos, product analyses - this matters.
The most analytically useful capability is get_growth with source='all' or a comma-separated list of sources. A single call returns growth rates for a keyword across Google, TikTok, Reddit, YouTube, Amazon, Wikipedia, and more - structured as a flat response where each source is a key with its own growth metrics.
This is useful for building leading indicator features. Consumer trend research consistently shows that TikTok and Reddit signals lead Google Search by 2-4 weeks, and Google Search leads Amazon purchase intent by another 1-2 weeks. A feature set that captures the cross-platform growth differential at each time step encodes this causal chain structurally - which outperforms single-source trend features in demand forecasting tasks.
Tools for this workflow
get_trendsRetrieve a full 5-year weekly time series for any keyword on any source - structured JSON with normalized values, absolute volume estimates, and data quality scores, ready for pandas or numpy.
get_trends(keyword='large language models', source='google', data_mode='weekly')
get_growthCompute period-over-period growth rates across multiple sources simultaneously - returns % change, start/end volumes, and direction for each source, useful as features in downstream models.
get_growth(keyword='large language models', source='google, reddit, youtube', percent_growth=['1M', '3M', '6M', '1Y'])
get_ranked_trendsPull ranked lists of fastest-growing topics on any platform - useful for generating keyword candidates for large-scale trend studies without manual curation.
get_ranked_trends(source='google', sort='yoy_pct_change', limit=50)
get_top_trendsRetrieve currently trending topics across platforms with no seed keyword - useful for time-stamped snapshots of trending content for longitudinal studies.
get_top_trends(type='TikTok Trending Hashtags', limit=30)
FAQ