How to build a trends dashboard
A trends dashboard is only as good as its data pipeline. Most attempts fail at the data layer - scrapers that break, single-source coverage, or relative-only values that cannot be meaningfully compared. This guide covers how to build a live trends dashboard using Trends MCP as the data source: what to query, how to structure the data, and how to display it with consistent cross-platform comparisons.
Free API access
Get your free API key
100 free requests per month. No credit card, no setup fee.
Replaced my manual Google Trends scraper in an afternoon. The data is clean and the latency is surprisingly low for a free tier.
We use it for keyword trend reports. The free monthly quota keeps us batching queries for weekly digests. Upgrading is there when we need more headroom.
Hooked it into my MCP server in like 20 minutes. The JSON response is well-structured and the docs are solid. Exactly what I needed.
We pipe weekly series into BigQuery for a few brand cohorts. Compared to maintaining our old Selenium job, this is boring in the best way. Uptime has been solid.
Great for slide-ready trend screenshots when leadership asks why we are prioritizing a feature. I wish the dashboard had saved views, but the API side is great.
Running it from Cursor with the MCP config took one try. I am not a trends person, but my side project now emails me when a niche keyword spikes hard week over week.
Using the growth endpoints to sanity-check retail names before I write up notes. Occasionally the normalization differs from what I see in the raw Google UI, but it is consistent run to run.
Pulling multi-source ranked lists into a notebook is straightforward. Error payloads are actually readable when I fat-finger a parameter, which matters more than people admit.
Does what it says. I knocked a star because onboarding assumed I already knew MCP wiring; a copy-paste block for Claude Desktop would have saved me 15 minutes.
We track TikTok hashtag momentum against paid spend in a Looker sheet. Not glamorous work, but it is the first tool my team did not argue about during rollout.
Retries are predictable and I have not seen weird HTML in responses (looking at you, scrapers). Would pay for a team key rotation flow, but for now we rotate manually.
Quick checks on retail buzz before we dig into filings. Not a silver bullet, but it is faster than opening twelve browser tabs and reconciling by hand.
Helpful for spotting whether a topic is a one-day meme or sticking around. I still cross-check with Search Console, but this gets me 80% of the signal in one call.
I demo this in workshops when people ask how to ground LLM answers in something fresher than training data. The MCP angle lands well with engineers who hate glue code.
Solid for client reporting. Billing is clear enough that finance stopped asking me what line item this is. Minor nit: peak hours can feel a touch slower, still acceptable.
I wired this behind a small CLI for contributors who want trend context in issues. Keeping the surface area tiny matters for OSS, and the schema has not churned on me yet.
Daily pulls for a 30-day window go straight into our internal scoreboard. Stakeholders finally stopped debating whose screenshot of Trends was newer.
We are pre-revenue, so free tier discipline matters. I hit the cap once during a brainstorm where everyone wanted to try random keywords. Learned to batch smarter.
Security review passed without drama: HTTPS, scoped keys, no bizarre third-party redirects in the chain we could find. That is rarer than vendors think.
I do not need this daily, but when App Store rank shifts look weird, having Reddit and news context in one place saves me from context switching across six apps.
I use it to see if a story is genuinely blowing up or just loud on one platform. It is not a replacement for reporting, but it keeps my ledes honest.
We moved off a brittle Playwright script that broke every time Google shuffled markup. Same data shape every week now, which is all I wanted from life.
Seasonal demand spikes line up with what we see in Amazon search interest here. Merch team stopped sending me screenshots from random tools that never matched.
Solid for client decks. I docked one star only because I still export to Sheets manually; a direct connector would be nice someday.
Steam concurrents plus Reddit chatter in one workflow beats our old spreadsheet ritual before milestone reviews.
Quick pulse on whether a feature name is confusing people in search before we ship copy. Cheap sanity check compared to a full survey.
Monitored from Grafana via a thin wrapper. p95 stayed under our SLO budget last month. One noisy day during a holiday but nothing alarming.
Narrative fights in meetings got shorter once we could point at the same trend line everyone agreed on. Sounds silly until you have lived through it.
Using normalized series as a weak prior in a forecasting experiment. Citation-friendly timestamps in the payload made reproducing runs less painful.
Approved for our pilot group after a quick vendor review. Would love SAML, not a blocker for our size.
YouTube search interest plus TikTok hashtags in one place helps me explain why a sponsor should care about a vertical without hand-waving.
Cron job hits the API before standup; Slack gets a compact summary. Took an afternoon to wire, has been stable for two quarters.
Useful for public-interest topics where search interest is a rough proxy for attention. I still triangulate with primary sources; this is one signal among several.
Runs in a VPC egress-only subnet with allowlisted domains. Fewer exceptions to explain to auditors than our last vendor.
Spotting when a topic is about to flood Discord saves my team from reactive moderation fires. Not perfect, but directionally right often enough.
For lean teams the ROI story writes itself. I would not build an in-house scraper for this anymore unless compliance forced it.
Examples in the docs match what the MCP actually returns. You would be surprised how rare that is in this category.
Pager stayed quiet. When something upstream flaked once, the error string told me which parameter to fix without opening logs first.
Students use it for coursework demos. Budget is tight so free tier matters; we coach them to cache aggressively.
Helps prep talking points when retail interest in our name swings after earnings. Not material disclosure, just context for Q&A prep.
Response sizes stay small enough for mobile hotspots. I hate APIs that dump megabytes for a sparkline.
What are you working on?
How will you connect?
A trends dashboard sounds simple: chart some trend lines, add growth percentages, ship it. The reality is that most early attempts collapse at the data layer within a few weeks. The scraper breaks. The data from different sources uses incompatible scales. The refresh runs into rate limits. This guide walks through building a dashboard that does not have those problems.
Start with the data architecture, not the UI
The most common mistake is starting with the visualization (choosing a chart library, designing the layout) before solving the data problem. Design the UI last. First, answer:
- What sources do you need to cover?
- What keywords or topics are you tracking?
- How frequently does the data need to refresh?
- How will you store and serve the data to the frontend?
Trends MCP as a backend simplifies questions 1-3 significantly. One connection gives your data pipeline access to Google, TikTok, Reddit, YouTube, Amazon, Wikipedia, news, web traffic, app downloads, npm, and Steam - all normalized to a consistent 0-100 scale. You still need to answer question 4 based on your stack.
Query architecture for a dashboard
Time series charts: get_trends
For the main trend line charts, use get_trends with data_mode='weekly'. This returns up to 5 years of weekly data per keyword per source. Query each keyword-source pair you want to display and store the results.
For most dashboards, querying multiple sources for each keyword gives you a richer picture. The response structure is consistent across sources - the same date/value/volume format regardless of whether you are querying Google or TikTok.
Practical note: One get_trends call covers one source. If you want to show Google, TikTok, and Reddit lines for the same keyword, that is three calls. Plan your daily query budget accordingly.
Growth scorecards: get_growth
Growth rate numbers (30-day, 90-day, 1-year) are the most-read data points on any trend dashboard. Use get_growth with source='all' to get cross-platform growth in a single call - this is the most efficient query pattern for dashboard scorecards.
The response includes percentage change, start/end volumes, direction (up/down/flat), and a data quality score for each source. The data quality score is useful for suppressing or flagging low-confidence values in your UI rather than displaying misleading growth figures.
Discovery / new keyword surfacing: get_ranked_trends
Most dashboards have a fixed keyword list and a discovery section. For the discovery section, run get_ranked_trends once per day on your primary source, sorted by wow_pct_change (week-over-week percent change) or yoy_pct_change. This surfaces keywords growing fastest that week, which you can present as "trending this week" cards and allow users to add to their tracked list.
Live trending / breaking signals: get_top_trends
For a "trending right now" section with no fixed keyword list, get_top_trends requires no seed keyword and returns what is actually trending at query time. This is the appropriate source for a live feed of breakout topics rather than monitored keyword lists.
Normalization and display
All Trends MCP values are normalized to 0-100 per platform. This is designed for cross-platform comparison on a shared axis. When plotting multiple sources for the same keyword, you can put them on the same chart without a second Y-axis or any transformation.
What you should not do is mix normalized values and absolute volume estimates on the same axis. If you display absolute volume, keep it to single-source charts. If you display normalized values, you can legitimately compare across sources.
See the normalization methodology page for a full explanation of how the 0-100 scale is calibrated across platforms.
Refresh strategy and caching
Daily refresh for time series: Run a scheduled job (e.g., cron at 2am UTC) that re-fetches the weekly time series for all tracked keywords. Store results in a database or object store. This keeps your charts current without continuous polling.
Real-time for live trending: get_top_trends can be refreshed every 15-60 minutes for a "trending now" section. The underlying data updates frequently for sources like TikTok and Reddit.
Cache growth scorecards for 24 hours: get_growth results change slowly - a 30-day growth figure on Monday and Tuesday is nearly identical. Serving cached results with a timestamp ("updated 6 hours ago") is accurate enough and reduces unnecessary queries.
Common mistakes
Mixing relative and absolute values: Google Trends' native 0-100 scale is relative to the peak in your query window. Trends MCP's 0-100 is normalized differently (consistently calibrated across time and sources). Do not mix these. If you are importing data from multiple sources and some comes from pytrends, the scales are not compatible.
Querying too many keywords on initial load: Dashboard load time degrades quickly if you query all keywords on page load. Pre-fetch and cache data server-side; serve stale-while-revalidate. Users should see charts instantly from cache, with a background refresh.
Ignoring data quality scores: Trends MCP returns a data_quality_score with each data point. A score below ~0.4 usually means sparse data or low platform coverage for that keyword. Displaying these values without flagging them misleads users. Either suppress low-quality data points or show them grayed out with a "limited data" indicator.
Single-source dashboards: If your dashboard only shows Google Trends, it is showing 1 of 15+ available signals. Users who rely on it for decisions are missing the 2-4 week lead time that TikTok and Reddit signals provide before Google Search catches up. Multi-source is not complexity - it is the minimum for a trustworthy trend picture.
Minimal viable implementation
The fastest path to a working trends dashboard:
- Set up a Trends MCP connection (one config entry in your AI client or one HTTP endpoint in your backend)
- Run
get_growth(keyword='your keyword', source='all', percent_growth=['1M', '3M', '1Y'])for your keyword list to populate scorecard numbers - Run
get_trends(keyword='your keyword', source='google', data_mode='weekly')for each keyword to build time series charts - Run
get_ranked_trends(source='google', sort='wow_pct_change', limit=20)for a discovery section - Store results, build the UI against the cached data
This four-query pattern covers the core of a production-grade trends dashboard. Add sources, additional keywords, and scheduled refresh once the baseline is working.
Connect
Add to your AI in 30 seconds
An API key is required to connect. Get your free key above, then copy the pre-filled config for your client.
Cursor
Cursor Settings → Tools & MCP → Add a Custom MCP Server
"trends-mcp": { "url": "https://api.trendsmcp.ai/mcp", "transport": "http", "headers": { "Authorization": "Bearer YOUR_API_KEY" } }
+ Add to Cursor
Or paste into Mac / Linux — ~/.cursor/mcp.json
Windows — %USERPROFILE%\.cursor\mcp.json
↑ Get your free key above first — the config won't work without it.
Claude Desktop
User → Settings → Developer → Edit Config — add inside mcpServers
"trends-mcp": { "command": "npx", "args": [ "-y", "mcp-remote", "https://api.trendsmcp.ai/mcp", "--header", "Authorization:${AUTH_HEADER}" ], "env": { "AUTH_HEADER": "Bearer YOUR_API_KEY" } }
Mac — ~/Library/Application Support/Claude/claude_desktop_config.json
Windows — %APPDATA%\Claude\claude_desktop_config.json
Fully quit and restart Claude Desktop after saving.
Claude Code (CLI)
claude mcp add --transport http trends-mcp https://api.trendsmcp.ai/mcp \ --header "Authorization: Bearer YOUR_API_KEY"
Windsurf
Settings → Advanced Settings → Cascade → Add custom server +
"trends-mcp": { "url": "https://api.trendsmcp.ai/mcp", "transport": "http", "headers": { "Authorization": "Bearer YOUR_API_KEY" } }
Mac / Linux — ~/.codeium/windsurf/mcp_config.json
Windows — %USERPROFILE%\.codeium\windsurf\mcp_config.json
Or: Command Palette → Windsurf: Configure MCP Servers
VS Code
Extensions sidebar → search @mcp trends-mcp → Install — or paste manually into .vscode/mcp.json inside servers
"trends-mcp": { "type": "http", "url": "https://api.trendsmcp.ai/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } }
Paste into .vscode/mcp.json, or:
Command Palette (⇧⌘P / Ctrl+Shift+P) → MCP: Add Server
Data Sources
What you can query
All data is normalized to a 0-100 scale for consistent cross-platform comparison.
Tools
What your AI can call
Four tools, organized by how you start. With a keyword, track history and growth. Without one, use discovery to see ranked movers or what is live right now.
You already have a keyword.
Chart how it moves over time and compare growth across sources.
No keyword required.
Ranked lists on one source with a growth sort you choose, or a live snapshot of what is trending across platforms.
Outputs
What you get back
FAQ