LangGraph and LangChain runtimes can treat trend pulls like any other tool: one MCP server, stable JSON, and explicit source strings for Google, YouTube, TikTok, Reddit, Amazon, news, npm, and Steam. The goal is fewer custom API wrappers and more time spent on routing, evaluation, and guardrails.
Agents that only read static files miss shifts that show up first in search, social video, commerce queries, or package downloads. A narrow tool surface reduces the odds that the model invents platform-specific parameters. Trends MCP exposes get_trends, get_growth, and get_top_trends so planners can branch on numeric change instead of parsing scraped HTML.
Name the required fields, repeat the exact source labels from the docs, and spell out when a keyword is not optional. TikTok expects hashtag style topics, Reddit expects a subreddit name without a prefix, and npm expects an exact package name. When live feeds are needed, require the feed type string exactly as listed for get_top_trends. The Model Context Protocol overview helps new readers understand why those constraints matter.
LangChain is rarely the only runtime. Batch jobs often call the REST POST endpoint while agents use MCP. The Python trends API page collects request shapes that mirror what an agent would call, which keeps evaluation notebooks aligned with production tools. For assistants and copilots that already enumerate tools, see MCP trend tools for assistants for field-level examples.
Cache repeated pulls within short windows, cap limit on live feeds, and log the returned dates so reviewers can trace summaries back to API output. When growth looks extreme, pull the matching get_trends window so the agent cites a series rather than a single point.
Create a key on trendsmcp.ai, register https://api.trendsmcp.ai/mcp with HTTP transport, and copy tool descriptions from the API reference.
FAQ