SpyTrend is an advertising intelligence platform that indexes Facebook, Meta and TikTok ads, and it has quietly become something more interesting than another ad spy tool: a large structured commercial dataset that autonomous AI agents can query directly in plain language. The vendor claims a corpus of 1.1 billion ads growing by 230 million new entries every month, and since 2026 it has shipped an official Model Context Protocol server so Claude, ChatGPT, Cursor, Codex and VS Code can reach all of it without an export step.
That combination is what makes the product worth a close reading rather than a directory listing. The competitive-research market is crowded and largely undifferentiated, so the two things that separate this platform from AdSpy or BigSpy are architectural: it groups ads by the operator behind them instead of listing creatives one at a time, and it exposes the whole database to an agent runtime over a single authenticated endpoint.
This article is a working audit rather than a recommendation. It covers what the platform actually indexes, how webmaster grouping and Pixel Intelligence work, what the twenty MCP tools return, what the plans cost, and — importantly — the five places where the vendor’s own pages contradict each other on price, coverage and billing. Everything below is drawn from the live site, its terms, its engineering blog and one independent editorial review, all read on 6 September 2026.
Table of contents
- What SpyTrend Is, and Who Built It
- Inside SpyTrend’s Webmaster Grouping and Pixel Intelligence
- SpyTrend MCP: The Ad Corpus Inside Claude, ChatGPT and Cursor
- What SpyTrend’s Own Test Runs Actually Found
- SpyTrend Pricing, Tokens and the Numbers That Do Not Agree
- SpyTrend vs AdSpy, BigSpy, AdHeart and SocialPeta
- Where SpyTrend Is Strong and Where It Is Thin
- The Terms, the Law and the SpyTrend Compliance Questions
- What SpyTrend Means for Marketing and IT Teams
- SpyTrend FAQ
- References and Further Reading
What SpyTrend Is, and Who Built It
The product is a subscription ad-intelligence service sold to media buyers, affiliate marketers, agencies and in-house performance teams. It sits in the same category as AdSpy, BigSpy, Minea and AdHeart, and it competes on a specific claim: that the useful unit of competitive research is the advertiser or the network, not the individual banner.
The corporate register
SpyTrend is operated by Yasha Limited, a Cypriot company registered under number HE 419811 at Office 908H, 9th floor, Nicolaou Pentadromos Center, Limassol 3312, Cyprus. The terms of service put the governing law in Cyprus and any litigation in the competent courts of Limassol. Payments run through Stripe on Visa and Mastercard, and the site is published in ten languages: English, Russian, Ukrainian, Spanish, Portuguese, German, French, Chinese, Japanese and Indonesian.
What the corpus contains
The headline figures are consistent across the homepage, the MCP page and the independent LaunchBuff review: 1.1 billion-plus ads in total, 230 million-plus new Meta ads added each month, and 14 million-plus advertiser fanpages carrying follower counts, founding dates, geographic mix and category tags. Ads are auto-classified into 33 top-level verticals and 343 subcategories, with no manual tagging required, and the vendor advertises a sub-second search response across the whole set.
The two sources, and the two that are missing
Coverage is Meta and TikTok. Both are queried in a single search with a source switch in the results feed, and the platform’s own FAQ is unusually direct that Google and native ad networks are not covered at all. The depth sits on the Meta side; webmaster grouping and the 230-million-a-month ingest run there, with TikTok positioned as a second source rather than an equal one.
| Attribute | Stated value |
|---|---|
| Operator | Yasha Limited (HE 419811), Limassol, Cyprus |
| Ad sources | Facebook / Meta and TikTok only |
| Corpus size | 1.1B+ ads, 230M+ new Meta ads per month |
| Advertiser records | 14M+ fanpages with followers, geo mix, categories |
| Classification | 33 verticals, 343 subcategories, automatic |
| Search response | Under 1 second, per the vendor |
| Paid entry price | $79/month Pro, $0 Starter tier |
| Agent access | MCP server at mcp.spytrend.com/mcp, OAuth 2.1 |
| Interface languages | 10, including Russian, Portuguese and Japanese |
Inside SpyTrend's Webmaster Grouping and Pixel Intelligence
Everything distinctive about the platform descends from one design decision: identity resolution across ads. Public ad libraries show you placements. This service tries to show you operators, and the machinery for doing that is the part competitors have not copied.
The four identifiers behind webmaster grouping
Webmaster Grouping links advertisements from the same operator by analysing four identifier types — Pixel ID, landing domain, Page ID and IP address. If one affiliate runs five domains behind a single pixel, the system presents all five as one advertising operation rather than five unrelated advertisers. The vendor’s about page claims SpyTrend is the only ad-spy service with the technology, and its own comparison table marks every named rival as lacking it.
Why a pixel is harder to hide than a page
The practical value shows up when an operator is churning assets. Fan pages get banned and replaced constantly, and from the outside the replacements look like new advertisers. The tracking pixel usually travels with the campaign bundle, so Pixel Intelligence follows the chain from ad to landing page to pixel to server and reassembles the operation. Hook a competitor’s pixel once and every new page raised under it surfaces on its own afterwards.
Days active as a profitability proxy
The platform’s “profitability detector” is simpler than the name suggests: it tracks how many days an ad has been live and treats five-plus days as a signal that the campaign is paying for itself. That is a heuristic, not a measurement — nobody outside the advertiser sees the return — but it is a defensible one, because unprofitable performance campaigns are usually killed inside a week. The recommended workflow leans on it: pick a vertical, open the Quick Win widget showing the five most active operators over seven days, filter to days active greater than five, then study what survives.
Deduplication, uncloaking and the rest of the feature set
Three supporting features matter more than the marketing gives them credit for. Deduplication collapses fifty copies of one banner into a single creative record with usage data attached. Landing page uncloaking follows redirect links to the real destination, which is the difference between a creative with a dead-end URL and a full funnel. And per-user categorisation lets a buyer tag ads and permanently suppress dismissed creatives from their own feed.
These are the vendor’s numbers about its own competitors, so read them as a marketing claim rather than a benchmark. What they do establish is the axis the company chose to compete on, and speed is at least a claim a buyer can falsify in a free trial within about ten minutes.
SpyTrend MCP: The Ad Corpus Inside Claude, ChatGPT and Cursor
The Model Context Protocol server is the reason this platform belongs in an AI publication rather than a marketing one. It turns a dashboard product into a data source that an agent can plan against, and it is a good worked example of where the protocol is actually landing commercially.
One endpoint, OAuth 2.1, no API keys
The connector URL is https://mcp.spytrend.com/mcp, and every supported client uses that same endpoint. Authentication is OAuth 2.1 with an explicit consent screen on each connection; the client opens a browser, the user signs in with their existing account, and no API key is ever pasted into a config file. The independent LaunchBuff review records the implementation as OAuth 2.1 with PKCE, which is what the current protocol specification requires of a remote server.
Client coverage, and the ChatGPT catch
Setup is a one-line command on most clients: claude mcp add --transport http spytrend https://mcp.spytrend.com/mcp for Claude, npx add-mcp for Cursor, codex mcp add for Codex and code --add-mcp for VS Code. ChatGPT is the exception and the vendor says so plainly — custom connectors need a Business, Enterprise or Edu workspace with Developer mode enabled by an administrator, so the option simply does not appear on a personal Free or Plus account.
The twenty tools, and the fourteen the page names
The catalogue is advertised as twenty tools spanning ads, advertisers, creatives, webmasters, hubs, shops, trends and media. The MCP page itself enumerates fourteen of them by name, which is a small documentation gap worth knowing about before you plan a workflow around a specific call. The named set breaks into four groups: five search tools, five drill-down tools, one rankings tool, and three account and favourites tools.
| Tool | Group | What it returns |
|---|---|---|
| search_ads | Search | Filtered ad list with geo, status, advertiser, landing domain, dates |
| search_creatives | Search | Creatives ranked by re-upload frequency |
| search_webmasters | Search | Networks with per-geo volume and dominant-geo ranking |
| search_advertisers | Search | Name typeahead with external ids and ad counts |
| search_hubs | Search | Destination profiles with an anti-cloaking filter |
| get_ad / get_advertiser | Drill down | Full ad card; advertiser counts, geo mix, domains, followers |
| get_webmaster / get_creative | Drill down | Domains, pixels, pages, geos; who runs one creative |
| get_media | Drill down | Signed media URLs for ads or creatives |
| get_trends | Rankings | Rankings by geo, advertiser, webmaster, app, timeline, scaling |
| add_to_favorites / list_favorites | Account | Folders for ads, creatives, webmasters and hubs |
| get_usage | Account | Plan, tokens used, tokens remaining — free to call |
What an agent workflow looks like in practice
The vendor’s own suggested pattern is the honest one: a cron job fires at 09:00, pulls yesterday’s scaling creatives in a target geography over MCP, hands the rows to a large language model to write a markdown brief, and drops the result into a team chat before standup. That is a few hours of work, not a project, and it is the same shape as the hosted connectors we looked at in the GMPlus Google Maps scraper MCP writeup — a data vendor deciding the agent is now a first-class client.
The same tools over plain HTTP
For teams that would rather not run an MCP client at all, the identical twenty-tool surface is available over plain HTTP with OAuth, with no SDK required and machine clients able to self-register. That matters for anyone wiring the corpus into an existing pipeline rather than a chat window, and it is a more open posture than the unified social data APIs that typically gate access behind issued keys and per-seat contracts.
What SpyTrend's Own Test Runs Actually Found
The engineering blog, last updated on 6 August 2026, publishes results from live queries against the production database. These are vendor-run tests and should be treated as such, but they are specific enough to be checkable by anyone with an account, which is more than most competitors offer.
De-anonymising a network from one parameter
The most striking figure in that post concerns tracking parameters. Taking a single value pulled from one operator’s links — sdmode=4 — and querying who else used it returned 2,367 other webmasters, plus a list of overlapping pixels and shared values. If the join is sound, that is an entire affiliate network, its resellers and one team’s scattered accounts surfaced from a single string.
Pixels, messengers and perceptual fingerprints
Three more numbers from the same run. One pixel ID returned 52 ads spread across several different fan pages, all pointing at one lander. A filter for ads sending traffic straight into a Telegram funnel returned 69,060 active placements, a segment that is effectively invisible in a normal feed. And an exact perceptual-fingerprint match — computer vision applied to banner images rather than to photographs — found 22,288 ads and 5,930 fan pages reusing one Brazilian gambling creative.
Scale detection with your own thresholds
The web interface offers preset growth buttons at 2x, 5x and 10x. Over MCP you set the threshold yourself — creatives that grew 4.5 times over three days, with at least eight new ads yesterday and six fresh accounts, all younger than three days. The vendor reports one visual going from zero to 246 ads inside a day under that detector, with daily launch history attached to each result.
Site-level reconnaissance
The database also holds destination-site data: monthly traffic, its sources, and the share that is paid. The published examples are a Brazilian gambling site at 613,000 visits a month with a 97% paid share and ad growth running ahead of traffic growth, a US camera-gear store at plus 1,013% traffic in a month, and a Brazilian-Venezuelan beauty store at roughly plus 866%. Whether those correlations generalise is exactly the sort of question a free trial is for.
SpyTrend Pricing, Tokens and the Numbers That Do Not Agree
Pricing is simple on the surface and genuinely confusing underneath, because three different vendor-controlled pages describe the token model in three different ways. Anyone budgeting agent workloads against this corpus needs to read all three before committing.
The three plans
Starter is free forever with limited database access and basic search. Pro is $79 a month for unlimited searches, full webmaster grouping, Pixel and Landing Intelligence, deduplication, data export, 90 days of history and one seat — plus one AI category included, with each additional category at $50 a month. Team starts from $63.20 per seat per month for 4 to 20 members with centralised billing and priority support; the site’s own worked example puts five members plus the owner at $355.50 a month after a 10% volume discount. Annual billing takes 20% off.
| Plan | Price | MCP tokens | Key inclusions |
|---|---|---|---|
| Starter | $0, free forever | 500 lifetime | Limited database access, basic search, no card needed |
| Pro | $79/month | 40,000/month | Unlimited searches, grouping, export, 90-day history, 1 seat |
| Extra AI category | $50/month each | — | One category is included with Pro |
| Team | from $63.20/seat/month | Per seat | 4–20 members, central billing, invite links, priority support |
| Enterprise | Contact support | Personal limit | Negotiated allowance |
How the token meter is described
On the MCP page the rule is that one token equals one delivered result — an ad or webmaster row in a search, a trends row, an entity card — so a default 20-result page costs 20 tokens. Creative media is dearer at 1 token per ad and 10 per unique creative, while navigation is free: the hub catalogue, similarity facets, favourites lists and the usage check itself. The vendor’s own arithmetic holds up, since 500 trial tokens at 20 per search is the “about 25 searches” the page advertises.
The same meter, described two other ways
The engineering blog says something different: that Meta analytics — search, de-anonymisation, trends and site statistics — costs nothing, and tokens are charged only for creative media and TikTok data. The LaunchBuff review describes a third arrangement: Meta rows at 1 token each and TikTok rows at 100 tokens each, Pro only. Those cannot all be true at once, and the gap between them is a factor of 100 on the workload most likely to burn an allowance.
Read the bottom bar honestly. A 500-token lifetime allowance is a sample gate rather than a research budget — it exists to let you confirm the data is real before paying, and the reviewer who tested the platform says exactly that. Anyone planning a standing agent workflow should assume Pro from the outset and model the TikTok multiplier before committing.
SpyTrend vs AdSpy, BigSpy, AdHeart and SocialPeta
The homepage carries a six-column comparison table, which is useful as a statement of intent and useless as an independent benchmark. It is still worth reproducing, because it shows precisely which claims the company is prepared to make in public.
The vendor’s own scorecard
By that table, SpyTrend and Tyver sit at $79 a month, BigSpy at $99, AdHeart at $130, AdSpy at $149 and SocialPeta above $2,000. Webmaster grouping, Pixel Intelligence, custom keyword tracking, per-user categorisation and GEO trends are marked as SpyTrend-only. Real-time status is shared with SocialPeta, with BigSpy marked partial, and vertical tagging is shared with SocialPeta and partially with AdHeart. The recurring competitive jab is the “3 to 7 day delay” the company attributes to BigSpy and AdSpy.
| Capability | SpyTrend | AdSpy | BigSpy | AdHeart | SocialPeta |
|---|---|---|---|---|---|
| List price per month | $79 | $149 | $99 | $130 | $2,000+ |
| Webmaster grouping | 4 ID types | No | No | No | No |
| Pixel Intelligence | Yes | No | No | No | No |
| Vertical tagging | 33 verticals | No | No | Partial | Yes |
| Custom keyword tracking | Yes | No | No | No | No |
| Real-time status | Yes | No | Partial | No | Yes |
| GEO trends | Yes | No | No | No | No |
Every row above is the vendor’s characterisation of its own rivals, reproduced here so the claims are visible rather than buried. Note too that Tyver, absent from the columns for space, is priced level with SpyTrend at $79 and marked as lacking every feature in the list.
SocialPeta is left off the chart because “$2,000-plus” is not a comparable figure — it is an enterprise contract, and putting it on the same axis would flatten every other bar into noise. On the corpus question the vendor concedes the shape of the field: it puts AdSpy at 177 million ads from major social platforms against its own 1.1 billion, but AdSpy indexes more networks than Meta and TikTok, so the two numbers are not measuring the same thing.
Reading a vendor comparison honestly
Three cautions apply to any table of this kind. Competitor prices drift and the entry tier quoted is rarely the tier a buyer ends up on. Feature ticks are self-assessed, and “partial” is doing heavy lifting in at least two rows. And the SpyTrend-only claims are architectural rather than absolute — rivals may resolve advertiser identity by other means without calling it webmaster grouping. Verify the two or three rows your workflow actually depends on.
Where SpyTrend Is Strong and Where It Is Thin
The independent LaunchBuff editorial review, published on 30 August 2026, scores the platform at 4.1 out of 5 — eight for ease of use, nine for value, nine for innovation, seven for feature depth and eight for support. That distribution matches what the product itself suggests: a sharp idea executed well, with a narrower surface than the category leaders.
The genuine strengths
Identity resolution is the real one. Grouping by pixel, domain, Page ID and IP produces a research object — the operator — that most competitors simply do not model, and everything downstream benefits from it. The MCP server is the second: a clean twenty-tool surface behind standard OAuth makes the corpus reachable from any compliant runtime without a custom integration or a negotiated data licence. Sub-second search and real-time status are the third, if they hold up under your own load.
The honest limits
Coverage is the big one. Meta and TikTok only means anyone buying Google, YouTube, programmatic display or native inventory needs a second tool for those channels, and the platform does not pretend otherwise. Feature depth scored lowest in the independent review. The 500-token Starter allowance is a demo, not a tier. And a single-seat Pro plan with 90 days of history is thin for an agency that wants year-over-year seasonality.
The trust surface
Some caution is warranted around the evidence the company publishes about itself. The testimonials are first-name-and-initial with a “results may vary” disclaimer. The “1100+ media buyers” figure is unaudited. The “running 5+ days means profitable” rule is a rule of thumb presented in the register of a metric. None of this is unusual for the category, and none of it is disqualifying — but it does mean the free tier exists to be used, not skipped.
The Terms, the Law and the SpyTrend Compliance Questions
The legal documents are more revealing than the marketing, and they contain the clearest internal contradictions on the site. Anyone procuring this for a regulated business should read the terms before the feature list.
What the terms actually commit to
Yasha Limited disclaims warranties of completeness, accuracy and timeliness, and caps total liability at the fees paid in the preceding three months. It offers a DMCA takedown route with a ten-business-day response commitment, and states that third-party advertising materials shown in the product remain the property of their owners and are displayed for analysis only. Disputes go to negotiation first, then to the courts of Limassol under Cypriot law.
Five places the site contradicts itself
Read across the pages and the same product is described inconsistently five times. None of these are fatal, and most look like documentation drift rather than bad faith, but each one is a question worth putting to sales in writing before a purchase order.
| Subject | One page says | Another page says |
|---|---|---|
| Money-back window | 7 days, on the homepage and pricing page | 14 days, in terms of service §5.1 |
| Free access | Starter is free forever, no card required | “Access to the Service requires a paid subscription”, §5.1 |
| Vertical count | 33 verticals, homepage and FAQ | 31 verticals, about page, updated March 2026 |
| Meta token cost | 1 token per delivered result, MCP page | Meta analytics is free, engineering blog, 6 Aug 2026 |
| Tool count | 20 tools, headline and API section | 14 tools actually named on the page |
Public data, GDPR and the scraping clause
The company’s position is that it processes only publicly available advertising information obtained through official ad transparency tools and public ad libraries, and it asserts compliance with GDPR and CCPA. That is the standard defence in this category and it is a reasonable one for ad creatives — but pixel identifiers, IP addresses and operator profiles sit closer to the personal-data line than a banner does, and a UK or EU buyer should satisfy itself under the ICO’s UK GDPR guidance rather than take the assertion at face value.
The clause that cuts both ways
There is an irony in the terms worth noting without overstating it. Section 4.2 forbids customers from using automated tools, bots or scrapers to access the service without written permission, and section 9.1 bars scraping, crawling or other automated data extraction from it, alongside a ban on reselling service data or building competing products. A platform whose entire value proposition is automated collection of other people’s public advertising reserves the right to prohibit exactly that against itself. It is also, for what it is worth, the normal shape of a SaaS agreement — and the sanctioned automated route is the MCP server the same company sells.
What SpyTrend Means for Marketing and IT Teams
Strip away the affiliate-marketing framing and there is a general lesson here for anyone building competitive intelligence into their operations, whether or not Meta ads are part of the plan.
Where it earns a seat
Three profiles get clear value. Performance teams buying Meta or TikTok inventory get operator-level research they cannot assemble by hand. Agencies get a defensible client deliverable — a map of a competitor’s advertising infrastructure rather than a folder of screenshots. And engineering teams building agent workflow automation get a structured, licensed corpus behind a standard protocol, which is far cheaper than negotiating raw data access.
Where it does not
If your paid media sits on Google, LinkedIn, programmatic display or retail media, this is the wrong tool and the vendor says so. If you need multi-year history, 90 days will not do. And if competitive research is an occasional need rather than a standing one, the free tier plus a manual pass through the public ad libraries will get you most of the way for nothing.
A sensible pilot
Run Starter for a week against one named competitor and check three things: does webmaster grouping actually connect the domains you already know they own, does uncloaking resolve their real landing pages, and does the search return in the time advertised. Then connect the MCP server to one client, ask for a single scaling-creatives brief, and read get_usage afterwards to see what the query really cost. Two hours of that will settle the token question the documentation does not. The same discipline applies to any external data analytics source you are about to make load-bearing.
The wider pattern
The broader signal is not about ad spying at all. It is that specialist data vendors are shipping MCP servers as a front door, priced per delivered row, so an agent can reach commercial datasets that used to require an integration project. Expect the pattern in every vertical where a proprietary corpus is the product, and expect the same three questions each time: what does a row cost, who is liable if it is wrong, and what happens to the query log.
SpyTrend FAQ
What is SpyTrend used for?
It is an ad intelligence platform for researching Facebook, Meta and TikTok advertising. Marketers use it to find competitors’ live campaigns, see which creatives have run long enough to be paying, trace landing pages behind cloaked links, and group ads by the operator running them.
Is SpyTrend free?
There is a free-forever Starter tier with limited database access and no card required, plus 500 lifetime MCP tokens. It is a validation allowance rather than a working plan. Paid access starts at $79 a month for Pro, with Team seats from $63.20.
Does SpyTrend cover Google or LinkedIn ads?
No. Coverage is Facebook/Meta and TikTok only, queried together with a source switch in the results feed. The vendor’s FAQ states explicitly that Google and native networks are not covered, so cross-channel teams need a second tool.
How does the SpyTrend MCP server work?
You point any MCP-compatible client at https://mcp.spytrend.com/mcp, authorise through OAuth 2.1 in a browser, and the agent can then call the catalogue’s tools in plain language. No API key is issued, and the connector grants the agent the same access as your own plan — nothing more.
What does a query cost in MCP tokens?
The MCP page prices one token per delivered result, so a default 20-result search costs 20 tokens against Pro’s 40,000 a month. Creative media costs more, at 1 token per ad and 10 per unique creative. Because the vendor’s blog and one independent review describe the meter differently, verify against get_usage on your own account.
Is using SpyTrend legal?
The company states it processes only publicly available advertising data from official transparency tools, and claims GDPR and CCPA compliance. Ad libraries are standard industry resources. Whether your specific use is lawful depends on your jurisdiction and what you do with pixel and operator data, so take advice rather than relying on a vendor FAQ.
Is SpyTrend better than AdSpy?
On the vendor’s own table it is cheaper — $79 against $149 — faster, and it offers webmaster grouping that AdSpy does not. AdSpy indexes more ad networks. If your buying is Meta and TikTok, the architecture argument is strong; if it is not, the coverage gap decides it for you.
References and Further Reading
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.