Developers who ship with Claude Code can attach normalized trend pulls from search, social, commerce, and news sources through one MCP endpoint, without wiring separate vendor APIs.
Claude Code is built for repo-aware edits, tests, and refactors. Market and product questions still show up in those same sessions: whether a dependency is gaining downloads, whether a game title is spiking on Steam, whether a keyword cluster is moving on Google Search, or whether a news topic is accelerating. A hosted Model Context Protocol server keeps those answers inside the tool channel the model already knows how to call, instead of opening brittle scrape workflows.
Trends MCP exposes three operations: historical series (get_trends), growth windows (get_growth), and ranked live feeds (get_top_trends). The source list spans Google surfaces, YouTube, TikTok, Reddit, Amazon, Wikipedia, npm, Steam, news volume, news sentiment, app downloads, and multiple live charts such as Google Trends, App Store leaders, and Spotify podcasts. One bearer token covers the stack, which matters when a developer is juggling side projects and does not want ten API dashboards.
Anthropic documents Claude Code MCP registration through the CLI. The hosted Trends MCP endpoint is https://api.trendsmcp.ai/mcp over HTTP transport, with Authorization: Bearer <token> on each tool call. After claude mcp add, verify the server appears in the local MCP list and run a dry prompt such as “Using TrendsMCP, show npm weekly downloads for react” so routing stays explicit. The phrase “Using TrendsMCP” matters because assistants sometimes default to web search when the tool name is vague.
Full reference material lives at https://trendsmcp.ai/docs. For readers who also use GUI clients, companion guides cover https://trendsmcp.ai/mcp-server-for-claude, https://trendsmcp.ai/mcp-server-for-cursor, and https://trendsmcp.ai/mcp-server-for-vs-code.
Before upgrading a major dependency, engineers often inspect release notes and GitHub issues. Adding quantitative demand from npm weekly downloads grounds the discussion. Ask for get_trends on the exact package string, then follow with get_growth across 30D, 3M, and 12M presets. If downloads are flat while issue volume rises, the story is different than a true adoption climb.
For studios or modders, Steam concurrent player series can show whether a patch coincided with renewed interest. Pull get_trends with the Steam source and a precise game title string, then compare against news volume for the franchise name if PR cycles might explain spikes.
When a CLI tool ships a paid tier, Amazon search demand and Google Shopping curves can show whether commercial intent keywords move together. Keep queries tightly scoped to product names to avoid noisy matches.
The free tier includes 100 requests per month without a card. Each get_trends or get_growth call counts once per source and keyword. get_top_trends counts per feed type and page. Batch curiosity can burn the quota; a useful habit is to draft the exact sources list before invoking the tool, then store outputs in the repo as structured notes if the team needs repeat access without new calls.
Some keyword and region combinations return data_unavailable when upstream pipelines are sparse. Reddit requires a subreddit name without the r/ prefix. TikTok expects hashtag or topic tokens that match how the catalog indexes volume. When a model proposes a vague keyword, narrow it before the call to save tokens and requests.
For ecosystem comparisons beyond Claude Code, see https://trendsmcp.ai/developer-ecosystem-trends. For REST shapes that mirror MCP payloads, see https://trendsmcp.ai/llm-trend-data-mcp-rest.
FAQ