· 12 min read
TrendTrack MCP: Connect Claude & ChatGPT to Your Data
TrendTrack MCP plugs your ecommerce research straight into Claude, ChatGPT and n8n. Here is the endpoint, all 11 tools, setup steps and prompts that work.
TrendTrack MCP is TrendTrack’s official Model Context Protocol server. It turns the platform from a dashboard you visit into a data source your AI assistant can query directly. Instead of filtering a table, exporting a CSV and pasting it into a chat, you paste one URL into Claude or ChatGPT, authorize once, and then ask for what you need in plain language.
The whole integration is one endpoint:
https://api.trendtrack.io/v1/mcpThat URL is stable. Set it up once and every new conversation after that already has TrendTrack attached — 11 tools, no client ID, no app secret, no JSON file to hand-edit. The OAuth manifest does the work.
This page covers what the MCP actually exposes, how to connect it in Claude and ChatGPT step by step, the other surfaces it plugs into (Perplexity, Manus, n8n, Cursor, raw REST), the prompts that people actually get value from, and the honest limits — including the one place where TrendTrack’s own pages contradict each other.
What Does TrendTrack MCP Actually Expose?
Four categories of data, all of it the same universe you already see in the dashboard, refreshed in under 24 hours:
| Data | What you can pull |
|---|---|
| Winning ads | Hook, creative, landing page, EU reach, lifetime, scaling verdict for any Meta ad |
| Scaling stores | Free-text store search, lookalike discovery, growth signals across the Shopify stores TrendTrack indexes |
| Email campaigns | Send cadence, subject lines, dominant content categories for any tracked brand |
| Hooks and angles | Top hooks running in a niche, angle patterns, media mix benchmarks |
The key difference from the dashboard is that an AI assistant can cross-reference those categories in one step. Ask a dashboard for “the hooks working in European skincare right now” and you get a filtered table. Ask the MCP and you get a written answer that already compared the creatives, counted the patterns and named the angle that repeats.
Two things it is not worth confusing it with:
- It is not the Chrome extension. The extension is a panel that reads the store you are looking at. The MCP is the full account-level dataset, reachable from anywhere.
- It is not a way to edit TrendTrack’s data. Discovery data — shops, ads, products, emails — is read-only through MCP. See the scopes section below for the one exception.
What You Need Before You Connect
Four things, and the third one is the usual reason a setup fails:
- A TrendTrack account. Free accounts can test the mechanics; heavier use needs a paid plan.
- Public API access enabled for your workspace. TrendTrack’s own ChatGPT documentation calls this out before anything else. If the toggle is off, the connector authorizes and then returns nothing.
- A client that supports custom MCP servers, on a plan that allows it. Claude custom connectors need a paid Claude plan. ChatGPT custom apps live behind Developer mode, which needs Plus, Team or Enterprise. Perplexity needs a paid plan. Cursor, n8n and raw HTTP have no such gate.
- About five minutes. The whole flow is paste-URL → log in → approve.
The TrendTrack MCP Endpoint and the 11 Tools
One endpoint, eleven tools, grouped into four jobs. This is the current documented set:
| Tool | Ask it | What comes back |
|---|---|---|
find_winning_products | ”Show me products scaling in pet supplies this week.” | Niche-ranked list with growth signals and the shop behind each one |
search_shops | ”Find Shopify stores selling sleep aids.” | Free-text store search across the TrendTrack universe |
find_similar_shops | ”Find shops that look like glossier.com.” | Lookalike list ranked by similarity, with the signals each one shares |
creative_inspiration_pack | ”Pull the best hooks and landing pages running in skincare right now.” | Top hooks, landing-page patterns, dominant angles, media mix benchmark |
brief_competitor | ”Give me a full brief on gymshark.com.” | Snapshot, top scaling ads scanned inline, email cadence, dominant angles, 5 opportunities |
scan_ad | ”Pull this Meta ad apart: <url>“ | Hook, creative, landing page, EU reach, lifetime, scaling verdict |
analyze_tracked_brand | ”Deep dive on alo.com — ads, hooks, landing pages, media mix.” | Brand overview, scaling ads, top hooks, LP gallery, format split |
analyze_shop_emails | ”What emails is huel.com sending? Frequency, subjects, vibe.” | Sending cadence, subject-line patterns, dominant content category |
daily_radar | ”What changed across my tracked brands since yesterday?” | Changelog: scaling ads, new launches, new landing pages, partnerships |
list_tracked_brands | ”What brands am I currently tracking?” | Every brandtracker with active ads, status and last update |
check_credits | ”How many credits do I have left this month?” | Balance, used this period, limit, reset date |
The first four are the discovery layer, the middle four turn raw signals into something paste-ready, the last three are monitoring and account plumbing. If you are evaluating the integration, find_winning_products, brief_competitor and daily_radar are the three worth testing first — they cover search, synthesis and monitoring in three prompts.
Worth knowing: TrendTrack’s ChatGPT documentation also references a
search_tiktok_librarytool for TikTok creative discovery, available through the same MCP server and mirrored by dedicated/v1/tiktok/libraryREST endpoints. TikTok coverage is additive to the Meta-first core — linked advertisers and the/v1/ads*endpoints remain Meta/Facebook only.
How to Connect TrendTrack MCP to Claude
Claude is the shortest path, because it reads the OAuth manifest on its own.
- Open Claude, then go to Settings → Connectors (desktop or web) and click Add custom connector.
- Paste the endpoint:
https://api.trendtrack.io/v1/mcp. No client ID, no secret, no JSON to edit — Claude reads the manifest from the URL. - Claude opens trendtrack.io. Log in, approve the scopes, and the tools appear in your conversation under the name Trendtrack.
You can revoke the connection from your TrendTrack account settings at any time. Combine this with the free TrendTrack Chrome extension and you have both ends covered: the extension for the store in front of you, the MCP for everything else.
How to Connect TrendTrack MCP to ChatGPT
ChatGPT uses the same URL and the same OAuth handshake, but the UI is hidden one level deeper:
- Click your profile in the bottom-left corner, then Settings.
- Go to Apps, then open Advanced settings.
- Enable Developer mode, then click Create an app.
- Name it Trendtrack, paste
https://api.trendtrack.io/v1/mcpas the MCP server URL, and click Create. - On TrendTrack, click Allow access to finish the OAuth flow.
TrendTrack is then available inside ChatGPT. Remember step 2 of the prerequisites: this only works with Developer mode, which means Plus, Team or Enterprise.
Other Surfaces TrendTrack MCP Plugs Into
The same server works anywhere that speaks MCP or plain HTTP:
| Surface | How it connects | Gate |
|---|---|---|
| Claude | Custom connector, OAuth | Paid Claude plan |
| ChatGPT | Custom app, Developer mode, OAuth | Plus / Team / Enterprise |
| Perplexity | Custom connector, paste URL, sign in | Paid plan |
| Manus | Custom MCP, URL or JSON import, OAuth | — |
| Gemini | No app UI yet; wire through Gemini CLI with one JSON entry, or Gemini Enterprise | — |
| Cursor | Standard MCP server entry | — |
| n8n | Ready-made template, paste your API key, then wire to Slack, Notion or Sheets | — |
| REST API | Direct HTTP with Bearer auth, no MCP required | — |
If you are not a chat-client person, n8n is the pragmatic choice. The pattern TrendTrack itself shows is a 9am cron that pulls the previous day’s scaling ads across your tracked brands, pipes them through an LLM for a markdown brief, and drops the result into a Slack channel before standup.
MCP or REST API? Which One to Use
Both doors lead to the same data, refreshed on the same schedule, and both draw on the same credits. The difference is who decides what to call.
| TrendTrack MCP | TrendTrack REST API | |
|---|---|---|
| Auth | OAuth in the browser, no key to store | Bearer API key |
| Who picks the tool | The model, from your sentence | You, in code |
| Best for | Chat clients, ad-hoc analysis, agent workflows | Scheduled jobs, apps, Make and Zapier, exact field selection |
| Predictability | High, but the model chooses the path | Deterministic — the same request returns the same shape every time |
A useful rule: use the MCP when a human is reading the answer, and the REST API when a machine is. If you want a Slack brief every morning, a cron hits the REST endpoints and formats the output — no model in the loop, no surprises in the payload. If you want to interrogate a niche on a Tuesday afternoon, the MCP is faster than writing a query.
One detail worth knowing if you mix them: API-key MCP callers are workspace-authorized exactly like REST API keys, and they skip the OAuth scopes entirely. So a service account wired into n8n does not need anyone to approve a consent screen — it inherits the workspace’s access level.
What You Can Actually Ask It
These are the prompts that produce something usable on the first try, because they name a niche, a store or a time window:
- Hook mining: “Find the hooks that worked in European skincare over the last three months and group them by angle.” You get recurring patterns ranked by impressions, not a list of links.
- Competitor teardown: “Give me a full brief on this competitor — what is working for them right now.” Returns the ad timeline, the strongest creatives, the angles, the landing pages and five opportunities.
- Creative brief: “Write a brief for a video ad in my niche based on the ten best-performing creatives right now.” This is the use case with the most obvious time saving: the brief is built from live ads instead of a template.
- Email teardown: “What emails is this brand sending, and what is the cadence?” Useful before your own welcome flow goes out.
- Morning diff: “What changed across my tracked brands since yesterday?” New launches, new landing pages, partnership announcements.
- Budget sanity check: “Check my remaining credits before you run a large query.” Cheap insurance once you start doing wide sweeps.
The quality gap is the point. An AI without market access answers from training data that may be months or years stale; the MCP answers from TrendTrack’s live index. That difference shows up immediately in campaign results, which is the claim you should test yourself rather than take on faith.
Read-Only, or Can It Write?
This is where TrendTrack’s own pages disagree, and it is worth knowing before you click approve.
The marketing FAQ still answers “Can I write to TrendTrack via the MCP?” with “No, read-only.” But the current connector documentation lists seven scopes and two of them are write scopes:
| Scope | Grants |
|---|---|
identity.read | Account, plan, credit balance |
discovery.read | Search shops, find winning products, scan ads |
brandtrackers.read | Your tracked brands, daily radar, email cadence |
brandtrackers.write | Add, rename, move and delete tracked brands and folders |
favorites.read | List favorite ads, shops, emails and folders |
favorites.write | Create, move, delete and share favorite ads, shops, emails and folders |
workspace_insights.read | Workspace-level signals and reports |
Read that as: TrendTrack’s underlying market data is read-only, but you can let the assistant manage your own workspace objects — brandtrackers and favorites. Destructive calls such as delete_favorite_folder require an explicit confirm: true.
Read the approval screen when it opens and only grant the write scopes if you actually want an agent reorganising your folders. Old setup guides that describe the integration as strictly read-only are describing an earlier version of the scopes.
Free Plan, Paid Plan and What MCP Costs You
TrendTrack’s own MCP page frames the split as: free gets the lookup tools with limited volume, to test the mechanic; paid unlocks the full tool set with unlimited volume, plus agentic workflows on a cron and Slack notifications. Treat that as the shape rather than the exact spec — the page itself notes the tiers are still being aligned with pricing, so confirm the current limits in your dashboard before you build a workflow on top of them.
Because queries consume account credits, the check_credits tool is not a gimmick. Wide sweeps (“compare every beauty store scaling in the EU”) burn credits fast; the twenty-second habit of asking for the balance first prevents the surprise.
If you are upgrading to run agentic workflows, the discount applies to every plan:
Apply the WINNING promo code for 20% off any TrendTrack plan →
WINNING takes 20% off Starter, Pro and Business for your first three months, on monthly or annual billing — the same ceiling as TrendTrack’s own codes TT02 and TT20. Our full TrendTrack coupon code breakdown has the tested table.
TrendTrack MCP vs the Chrome Extension
People keep asking which one they need. They answer different questions:
- The extension is a teaser on the current page — you are on a store, and it tells you about that store. Zero setup, free.
- The MCP is the full stack, agent-ready. You can cross-reference brands, generate briefs, and automate the whole loop.
Most people end up with both, because the moments are different: the extension for browsing, the MCP for research and reporting. If you only research one competitor at a time, the extension plus the step-by-step TrendTrack walkthrough may be enough. The MCP earns its keep when you are tracking a set of brands and want output on a schedule.
TrendTrack MCP FAQ
Is TrendTrack MCP free?
The connection itself costs nothing extra — it is part of your TrendTrack account, and there is no separate MCP fee. What varies by plan is tool access and volume: TrendTrack’s own comparison gives free accounts the lookup tools with limited volume for testing, and puts the full tool set plus unlimited agentic workflows on paid plans. Your client matters too, since Claude custom connectors and ChatGPT custom apps both require a paid plan on that side.
Does TrendTrack MCP work with Cursor, n8n, Make and Zapier?
Cursor works, because it speaks the MCP standard natively. n8n works through a ready-made template where you paste your API key and wire the output to Slack, Notion or Sheets. Make and Zapier are not MCP clients, so they connect through the REST API over plain HTTP instead — the same data, a different door.
Can TrendTrack MCP write to my account?
Partly, and this changed. The live connector documentation grants write scopes for brandtrackers and favorites, so an assistant can add, move, rename and delete your tracked brands and organise or share your favorite folders — with destructive actions requiring explicit confirmation. TrendTrack’s underlying discovery data stays read-only. Older pages that call the whole integration read-only are describing the previous scope set.
The Short Version
TrendTrack MCP is the most useful thing the platform has shipped for anyone already using an AI assistant: one stable endpoint, eleven tools, OAuth instead of API keys, and the same dataset the dashboard reads. Connect it in Claude if you want the fastest setup, in ChatGPT if you are already paying for it, or in n8n if you want a brief waiting in Slack every morning.
The honest limits: data refreshes in under 24 hours rather than by the second, large queries consume credits, discovery data stays read-only, and both Claude and ChatGPT put custom connectors behind a paid plan. None of that is a dealbreaker — it just means the integration rewards people who know what they want to ask.
If you are still deciding whether the platform is worth the subscription at all, start with the full TrendTrack review and the pricing breakdown after the discount. If you have already decided and just want to pay less for it, code WINNING takes 20% off every plan for your first three months — and it is the only TrendTrack coupon code worth typing at checkout.
