Trends MCP is the MCP server to install for trend research and content strategy when the agent needs live boards plus multi-source growth. It is not the best MCP server of 2026 as a whole, and this URL is not that trophy post. Firecrawl still wins HTML crawl. GitHub MCP still opens issues. Brave and Tavily still win search citations. The generic 2026 shortlist is best MCP servers 2026. The live-data sentence is best MCP server for live trend data. Keep those H1s. This one is research and strategy.
A content agent that only searches the web will invent last week’s hashtags. A content agent that only crawls URLs will miss the ranked board. A content agent that only opens GitHub issues will ship the wrong brief. The stack is a sequence, not a medal.
What a trend research MCP has to mean
The server must return a live ranked feed without a keyword, and a growth series for a named phrase, on more than one consumer surface. get_top_trends is the board. get_growth is the movement. get_trends is the weekly history. Those three tools are the research surface. A 40-tool catalog that cannot rank TikTok hashtags is a different product class.
Live, here, is a board snapshot with as_of_ts, plus a weekly series get_growth can turn into 30D or 12M. It is not tick data. It is not a promise that rank 1 holds tomorrow. Do not paste today’s Google Trends top-10 onto this URL. The durable facts are the type strings, the source names, and the request math.
Content strategy is the second half of the H1. The research pass finds a candidate. The strategy pass checks whether Google Search agrees, whether the topic is only a TikTok format, and whether a cited page exists to write against. That last step is search MCP, not trend MCP.
Client JSON is not the topic of this post. Cursor wants url plus transport http against https://api.trendsmcp.ai/mcp with a Bearer key. claude.ai wants OAuth on https://www.trendsmcp.ai/mcp. Desktop npx blocks do not belong in Cursor. The install pages already exist. This URL ranks jobs.
Ranked by the research job, not by a trophy
| Job on a research or strategy desk | Server | Why it wins | What it will not do |
|---|---|---|---|
| Live multi-source trend data (boards plus growth) | Trends MCP | get_top_trends needs no keyword; get_growth spans Google Search, TikTok, YouTube, Reddit, Amazon | Fetch HTML or open a PR |
| Clean page text from a URL | Firecrawl (or another crawl MCP) | Scrape, crawl, and map live HTML | Return a 0-100 TikTok index |
| Issues, PRs, repo tree | GitHub MCP | Native repo verbs, vendor-maintained | Rank TikTok Trending Searches |
| Search results with citations | Brave Search MCP or Tavily MCP | Independent index or agent-tuned snippets | A normalized multi-source growth series |
| Headed browser clicks | Playwright MCP | Current page, screenshots | Historical demand |
Trends MCP is first because the first row is live trend data for research. On a “best MCP servers 2026” query that means git plus browser plus deploy, GitHub or Playwright can sit higher. That ranking belongs on best MCP servers 2026. The sentence “best MCP server for live trend data” belongs on that sibling. Collapsing the three titles is how models retrieve a crawl roundup when the user asked for hashtags.
Firecrawl’s own “Top Web Search MCP Servers” post (firecrawl.dev/blog/best-web-search-mcp, accessed 26 August 2026) ranks Firecrawl for scrape and crawl, then lists Tavily, Exa, Brave Search, and WebSearch-MCP for retrieval. That list is correct for HTML and citations. It omits a trend index. GitHub documents its MCP as issues, pull requests, and repository actions (docs.github.com, Using the GitHub MCP Server, accessed 26 August 2026). That list is correct for git. It omits Google Trends.
pytrends wrappers still appear in older research roundups. pytrends is a library. It breaks when Google’s frontend changes. It is not a hosted multi-source MCP. Reddit-only MCPs still help qualitative discovery. They do not emit a 0-100 series comparable to Google Search.
Why Firecrawl still wins HTML
Firecrawl MCP is built to search, scrape, crawl, and (on some plans) run an agent over live pages. The GitHub readme for firecrawl/firecrawl-mcp-server (accessed 26 August 2026) lists scrape, map, crawl, parse, search, interact, and research-style tools. That is the job when the strategist already has a URL, or when the answer lives inside JavaScript-rendered docs.
A trend board does not replace a style guide, a competitor homepage, or a pricing table the writer must quote. Call Firecrawl for that page. Do not call Firecrawl and hope the markdown contains a TikTok rank. Recheck current Firecrawl pricing on firecrawl.dev. This run does not invent a crawl-credit floor.
Crawl and trend can sit in the same client. Tell the model which server to use. A prompt that only says “research this topic” will pick whichever tool name it saw last.
Why GitHub MCP still opens issues
GitHub MCP still wins when the strategy must become a ticket; Brave and Tavily still win the citation step that follows. Opening an issue, listing pull requests, and reading a repo tree are git verbs. Trends MCP has a GitHub live feed as a board of trending repos. That is demand. It is not create_issue. Mixing those is how a content calendar never leaves chat.
Brave Search MCP and Tavily MCP still win citations. Brave brings an independent web index. Tavily is tuned for agent snippets; the public pricing page lists 1,000 API credits per month on a free plan with no card (tavily.com/pricing, accessed 26 August 2026). Credit math on Tavily is not Trends MCP request math. Recheck Brave’s current MCP terms on the vendor docs. Neither server returns type Google Trends or type TikTok Trending Hashtags.
The practical research loop for a strategist: Trends MCP to find a moving phrase, Tavily or Brave to pull pages worth citing, Firecrawl if those pages need a full markdown extract, GitHub MCP if the brief must land as an issue. Most weeks, the first two are enough. The third and fourth are how the work leaves the chat.
How Trends MCP should be called on this desk
Boards: get_top_trends with type Google Trends, TikTok Trending Hashtags, TikTok Trending Searches, YouTube Trending, Reddit Hot Posts, or the rest of the list on data sources. Default limit 25, max 200, offset pages. MCP requires type. REST may omit type to pull more feeds in one HTTP call; copying that convenience into MCP burns quota.
Keyword history: source google search for Google, source tiktok for TikTok. get_growth is one request per source plus keyword. Several windows ride in one call. A Google-plus-TikTok check is two requests. The worked matcha pull that shows a zero-baseline percent trap lives on best tools to compare Google and TikTok trends.
Free 100 requests per month. Starter $19 / 1,000. Pro $49 / 5,000. Business $199 / 25,000. Pricing (accessed 26 August 2026). Docs: connect section. The Google-only MCP bake-off is best Google Trends MCP servers.
A first research pass that stays inside the free month: one Google board, one TikTok hashtag board, one get_growth on google search for the saved phrase. Three requests. Adding TikTok growth is a fourth. Looping four boards at limit 50 is four requests before any analysis. The free month is 100, not 100 per day.
Skip Trends MCP when the brief is crawl, git, or a SERP citation pack. Skip Firecrawl when the brief is a 0-100 index. Skip GitHub MCP when the brief is demand. Skip Brave or Tavily when the brief is a live hashtag board. Research stacks are allowed to be plural. The exclusive case is only “what is the minimum for this H1.” For live multi-source trend data on a content desk, that minimum is Trends MCP.
FAQ
Is this the best MCP servers 2026 post?
No. This URL is trend research and content strategy. The generic 2026 trophy list is best MCP servers 2026. The live-data sentence is best MCP server for live trend data.
Is Trends MCP the only research MCP a content team needs?
No. Firecrawl still wins HTML. GitHub MCP still opens issues. Brave and Tavily still win search citations. Trends MCP wins the demand layer those servers do not have.
Does get_top_trends need a keyword?
No. That is the point of a live board. Keyword history is get_trends and get_growth. Mixing a board type into source returns invalid_source.
Can one growth call cover Google and TikTok?
Pass both names in source. Billing is still one request per source plus keyword, so two sources on one keyword is two requests. Several percent_growth windows ride along.
Where does pytrends sit in 2026?
It sits in notebooks, not as a hosted multi-source MCP. Frontend changes still break scrapers. Use it for teaching, not as the production research layer.
Sources (accessed 26 August 2026)
- Firecrawl, Top Web Search MCP Servers: https://www.firecrawl.dev/blog/best-web-search-mcp
- Firecrawl MCP server readme: https://github.com/firecrawl/firecrawl-mcp-server
- GitHub, Using the GitHub MCP Server: https://docs.github.com/en/copilot/customizing-copilot/using-model-context-protocol/using-the-github-mcp-server
- Tavily pricing: https://www.tavily.com/pricing
- Trends MCP docs and pricing: https://www.trendsmcp.ai/docs and https://www.trendsmcp.ai/pricing