Real-time trends API for AI agents

Most trend APIs return yesterday's data. Trends MCP delivers live, real-time trend signals from Google Search, TikTok, Reddit, YouTube, Amazon, and 8 more platforms - structured for AI agents via the Model Context Protocol. One endpoint. No polling. No scraping.

Updated 2026-08-19

A real-time trends API for agents is a live feed plus a history API, not a screenshot of yesterday's dashboard. Trends MCP exposes get_top_trends for keyword-free live lists (Google Trends, TikTok hashtags, Reddit Hot Posts, Wikipedia Trending, SimilarWeb Top Websites, Amazon bestsellers, Google News RSS, Google Play, and related feeds) and get_trends / get_growth for the 0-100 series behind a known phrase. One MCP connection covers 25+ sources. History depth is about 5 years weekly. The free plan includes 100 requests per month. Starter is $19/month.

Live feeds do not take a keyword

get_top_trends is the real-time entry point. The caller chooses a feed type and a limit. There is no seed phrase. That is the difference from get_trends, which always needs a keyword and a source. Discovery workflows start with a feed, then pass surviving phrases into growth and history tools.

Feed types are platform-shaped: TikTok Trending Hashtags, Reddit Hot Posts, Wikipedia Trending, SimilarWeb Top Websites, Amazon Best Sellers Top Rated, Google News RSS, Google Play, Google Trends. The list is not a promise that every social network on earth is included. It is the catalog the MCP server actually serves.

Polling the same feed every 30 seconds wastes monthly quota and rarely changes the ranking. News and TikTok justify shorter intervals during a breaking window. Weekly research still starts with history tools, not with a live feed.

History tools stay on the same connection

After a phrase appears on a live feed, get_trends answers whether it has a 5-year shape or is a one-day flare. get_growth with 7D, 1M, and 3M on source='all' or a short source list is the usual confirmation. Daily data_mode exists for the last 30 days when the story is still moving.

The 0-100 index is shared across sources so an assistant can compare Google Search and YouTube without a second vendor. Absolute volume is optional and directional. Live feed items may not carry the same volume field as history points; agents should treat feed rank as order, then confirm with a history call.

REST clients POST to https://api.trendsmcp.ai/api with mode: "top_trends" for feeds and with source plus keyword for history. MCP clients invoke the tools directly. Google Trends API documents the search-volume variant of the same contract.

How this differs from native platform APIs

Google, TikTok, Reddit, YouTube, and Amazon each ship APIs for their own objects: ads, listings, videos, posts. None of them offer a single agent session that returns live breakout lists from the others, already normalized. Building that in-house means keys, scrapers, and a join layer that breaks when any frontend changes.

Trends MCP does not replace those native APIs for write actions, ads delivery, or owned-account analytics. It replaces the research join: live feeds plus history for 25+ sources behind one token.

SERP and social-firehose vendors sell other slices of "real time." Teams should check current vendor pricing for firehoses billed by document. Trends MCP is billed as research requests per month, which matches assistant workloads that call a feed, then a handful of keywords, then stop.

Who needs live feeds versus weekly series

News desks and social producers need get_top_trends during the day. Ecommerce and SEO teams can live on weekly get_growth plus an occasional feed check so they do not chase a hashtag that will be gone by the time a landing page ships. Investor chatbots often mix a morning feed pull with stored watchlist history.

Daily trending reports and the Google Trends API are the adjacent entry points. This page is the real-time contract: feeds without keywords, history with keywords, one MCP session.

Freshness, caching, and limits

"Real time" is not a single SLA. Search breakout lists lag the raw query stream. Wikipedia and news can move faster than Google's public Trends UI. A client that needs sub-minute certainty still belongs on a firehose product.

Trends MCP does not require the caller to run a scrape loop. The server fetches the feed when the tool runs. Clients should still cache identical type plus limit calls inside a few minutes if several agents share a desk.

Quota is 100 requests per month on the free plan, $19/month on Starter. Each successful feed pull counts. A 24/7 poller will exhaust the month. Scheduled pulls (hourly for one feed during an event, daily otherwise) fit the cap.

Low-quality or empty feeds return short lists or errors rather than invented topics. Assistants should not pad a brief with placeholder "trending" rows when the tool returns nothing.

A workable daily ritual is one feed pull in the morning, a short human or model filter, then at most ten get_growth calls on phrases that survived. That is a dozen requests, which fits the free month if it does not run on weekends as well, and it leaves room for ad-hoc history during the week. Turning every Slack mention into a live feed call is how desks hit 429 errors before Friday.

Comparison: live trend APIs

Trends MCP is one MCP connection covering 25+ sources. Live discovery is get_top_trends with no keyword. Confirmation is get_trends and get_growth on a 0-100 index with about 5 years of weekly history. Free volume is 100 requests per month.

Official platform APIs remain the path for each company's own write and ads objects.

Single-source trend scrapers and SERP APIs remain available for teams that only need Google. Teams should check current vendor pricing. Trends MCP is the multi-source live-plus-history API for assistants, not a replacement for a social listening firehose.

get_trends

Retrieve a structured time series for any keyword at request time. Each point includes date, a 0-100 value, and optional absolute volume for agent pipelines.

get_trends(keyword='electric vehicles', source='google search', data_mode='daily')

get_growth

Return cross-platform growth metrics in one call, including short windows such as 7D and 1M that sit next to 3M research windows.

get_growth(keyword='electric vehicles', source='all', percent_growth=['7D', '1M', '3M'])

get_ranked_trends

Query a ranked, precomputed list of the top trending keywords right now - optimized for real-time latency, returns the highest-signal topics without requiring a seed keyword.

get_ranked_trends(source='google search', sort='wow_pct_change', limit=50)

get_top_trends

Pull the live trending feed from a platform with no keyword. This is the primary real-time discovery call.

get_top_trends(type='TikTok Trending Hashtags', limit=30)

Common questions

Freshness varies by source. Google Trends-style search feeds typically refresh within hours. TikTok and Reddit live feeds are near-live, on the order of minutes to hours. Wikipedia and news feeds update as those systems publish. History tools (get_trends, get_growth) read stored weekly or daily series rather than scraping a page on each click. Live ranking tools (get_top_trends) request the current feed at call time.
Most trend APIs expose one platform, one schema, and a separate key. Trends MCP is one MCP connection covering 25+ sources with a shared 0-100 index for history calls. Assistants receive JSON they can rank, filter, and compare without maintaining a scraper fleet.
No. Trends MCP holds source access. A client registers one MCP endpoint or one REST Bearer token and can call Google, TikTok, Reddit, YouTube, Amazon, and the rest of the catalog in the same session.
get_trends returns a keyword time series. get_growth returns period-over-period change, including cross-source lists. get_top_trends returns live feeds and does not require a keyword. A compatibility ranked-list helper may appear in older docs; new workflows should prefer those three tools.