Chat with your data over MCP

Filter brands by niche, monthly visits, Meta or Google ad activity, storefront status, e-commerce platform, product count, apps, payment methods, or technologies.

Random sample of trending brands. The right tool for "show me brands" / "give me examples".

Full details for a brand by domain (e.g., gymshark.com).

Full details for a brand by full website URL.

Batch fetch up to 100 brands by ID.

Free-form search by brand name, domain, or social handle.

Batch look up to 50 identifiers in one call.

Same as search_brands but takes a JSON filter body. Use when filters get long.

Profile + active Meta ads + products in a single call. Best for "tell me about X".

The Facebook/Meta pages running ads for a brand.

Ads or posts for a specific brand on meta, tiktok, or instagram.

Search every Meta ad we track by copy, spend, reach, targeting, dates.

Search every TikTok post by description, engagement, duration.

Search every Instagram post by caption, hashtags, engagement.

JSON-body counterpart of search_meta_ads.

JSON-body counterpart of search_tiktok_ads.

JSON-body counterpart of search_instagram_posts.

Single Meta ad by ID.

Single TikTok post by ID.

Single Instagram post by ID.

Lookup a Meta ad by its Meta Ads Library numeric ID — the one in facebook.com/ads/library/?id=... URLs.

Cross-platform resolver: paste any Meta Ads Library, TikTok video, or Instagram post/reel URL — or a Brandsearch share link (app.brandsearch.co/ad/…) — and get back the matching ad/post (with a platform field for routing).

The ad's creative, inlined as a picture the model can actually see — so it can describe, compare, and critique creatives rather than just link to them. Takes platform (meta / tiktok / instagram / email), id, and media. A video ad comes back as its cover frame; media="video" instead reports the video's type, size, and a download URL. With platform="email" it returns the email's screenshot, so the AI can read the whole email.

AI creative breakdown of one ad or email — concept, angle, hook, audience, funnel stage, suggested tests. Costs AI credits (19 for a video, 14 for an image or an email); a previously generated analysis comes back free.

What is said in a Meta video ad — timed speech segments, the opening hook line, the spoken language, and the duration. An existing transcript comes back straight away; otherwise one is produced, which can take a minute or more. Costs 1 credit per transcript returned. Image ads are refused with ad_not_video. See transcribe a video ad.

Random high-quality Meta ads.

Random viral TikTok posts.

Random viral Instagram posts.

Random sample of marketing emails from trending brands — the default for open-ended requests like "show me email inspiration".

Search every email we track by subject and preview text, type, send date, and brand.

JSON-body counterpart of search_emails.

One brand's emails, newest first.

Single email by ID. Pair it with download_ad(platform="email") to see the email itself.

Email count, type mix, the 50 brands sending the most, and emails per day and per weekday, over any slice.

Six-dimension Meta ads dashboard for a single brand (media mix, funnel, CTAs, demography, spend).

TikTok or Instagram dashboard for a single brand.

Meta ads dashboard across all brands matching a filter slice.

TikTok dashboard across all brands.

Instagram dashboard across all brands.

Paginated list of the user's folders for a given folder_type.

One folder's metadata and the IDs of its saved items. Ads folder items leave out the REST route's is_valid / scraped_at, which say nothing about whether an ad can be fetched: an item needs no processing first, and get_folder_ads looks each one up by its ad_id or, failing that, by its saved link. For mobile_feed, the folder alone; its posts come from get_mobile_feed with folder_id.

Resolve an ads or spectre folder to full ad records (Meta / TikTok / Instagram, and emails saved to an ads folder). spectre folders also accept max_ads_per_brand, status, is_video, is_image, sort_by, sort_order.

Resolve a brands or spectre folder to full brand records. Every saved item comes back as one row; the ones we can't match to a brand are flagged with _swipe.resolved: false instead of being dropped.

Get a board and its linked directories by ID.

Create a new folder for the user. Takes folder_type and a name (1-100 chars; 1-50 for mobile_feed, which also takes a platform: instagram by default, or facebook).

Save an item to a folder. Per kind: ads → ad_id + type (one of meta/tiktok/instagram) for an ad found through any other tool, or url + type for one that isn't in our catalogue; brands → brand_id (a domain; a Meta page ID also works); spectre → brand_id or the brand's url on its own, or url + page_id for Meta Ads Library mode. mobile_feed → post_id, a post from get_mobile_feed. Idempotent: re-saving the same item just attaches it to the additional folder and returns already_in_folder: true.

Rename a folder. Takes folder_type, folder_id, and a new name (1-100 chars; 1-50 for mobile_feed). Works on a folder the user owns or has edit access to; a duplicate name returns swipe_folder_name_taken.

Delete a folder (owner only). Takes folder_type and folder_id. Removes the folder and its item links; the saved ads/advertisers/brands themselves are kept, and a mobile_feed folder's posts stay in the feed.

Remove a saved item from a folder without deleting the item itself. Takes folder_type, folder_id, and item_id (the saved_ad_id / saved_advertiser_id / tracked_brand_id from get_folder, or a post's id for mobile_feed). Idempotent: removing an absent item returns already_removed: true.

The user's posts, newest received first, with the same filters as the API: platform, type (reel / post), media_format, folder_id, q, received_from, received_to, received_after, sort_by, sort_order. Free.

Take a post out of the user's feed, like Remove from feed in the app. Takes post_id.

Search the Product Library across every brand: one card per market (all the stores selling the same product count once). Filter by text (q), niche, product_type, source (shopify = has ads, tiktok_shop), price, stock, ad counts, first/last advertised dates, est_revenue_min_usd, sellers_min / sellers_max and market_country (EU and UK only). Sort by est_revenue (the default), relevance (the default when q is given), newest, ads, active_ads, ads_30d, daily_spend, sellers, price_asc / price_desc or tiktok_shop_plays. Each card's revenue is a monthly band, and revenue.scope says whether it is the store's or the whole market's. 1 credit per product returned.

One product in full by the id of a product card: price, ads, market, the revenue band, plus its audience, the top 5 stores selling it, its TikTok Shop footprint and whether a supplier match exists. 1 credit.

The same, from a storefront product URL pasted as copied (query string, www., collection or locale prefix all resolved), including a Meta ad's landing page (creative.link_url, shown as Landing page: in the ad's summary). 1 credit.

Up to 20 product URLs at once: found product cards in the order of the URLs, not_found the misses. 1 credit per product found.

The Meta ads promoting a product and the stores we matched as selling it (as on its Ads tab in the dashboard), newest first, 10 per page (5 pages at most). Each ad's id works with get_meta_ad and download_ad. 1 credit per ad returned.

Every store selling a product, most active ads first, at most 50: price in USD (none for a store whose listing we cannot read), ads, and which row is the product's own store. 25 per page; 1 credit per store returned.

A product's ad activity day by day over the last days days (1-90, default 30): the store's ads (total, active, started in the last 30 days, estimated daily spend) and its market's (stores advertising, estimated daily spend). Shows whether its advertising is growing or slowing, not its sales. 1 credit; a product with no recorded activity returns an empty list, free.

The 1688 supplier matched to a product and its unit economics: unit and bulk cost, MOQ, estimated shipping, gross margin (before shipping and payment fees), markup, break-even ROAS and CPA (which take them into account), a verdict and up to 3 alternative listings. Only for products whose sourcing_available is true. 1 credit; a product with no match costs none.

How a product's market advertises it, from the AI analysis of its Meta ads: per facet (hook, angle, funnel stage, awareness level, pain, audience, motivation, tone, strategy pattern), the five most common values and their shares. 1 credit; a product none of whose market's ads has been analysed (most of the library), or whose analyses show nothing, costs none.

Who a product's market reaches with its EU/UK Meta ads, as shares: women and men, each age band by gender, and the 15 leading countries. 1 credit; a product whose market disclosed no reach (about half the library) costs none.

Products that look like a given one (image similarity), from other markets, closest first; limit 1-20 (default 10). 1 credit per product returned.

The product's picture itself, inline, so the AI can look at it — fetching image_url directly usually fails. size is compact (default, at most 800 px wide) or full. 1 credit; a product whose image we do not hold costs none.

Every store selling one product — its market — best first, 50 per page with a cursor. Pass a product card's market.id. history=true adds the market's daily series (advertisers, estimated spend). 1 credit per seller returned.

One brand's own catalogue (to find products across brands, use search_products). product_type can be all, bestsellers (ranked, best first) or latest (newest first). created_from / created_to keep only products added to the store in that window — handy for "what has this brand launched lately?".

Canonical reference values (e.g., apps, technologies, niches, country-codes). Use when you need to know the valid values for a filter.

Current API usage + remaining daily/monthly quota.

The account's AI-credit balance — what analyze_ad spends from.

Which account the connection is bound to — user, email, the API key in use, team leader, quota ceilings, the permissions granted to the session, and how many folders of each type it can see. The first thing to ask when the AI sees different folders than the dashboard does. If it reports account_link_broken, your folders are safe but unreachable: disconnect and reconnect the Brandsearch connector to fix it. Unmetered.

Latest timestamps for brands, Meta ads, TikTok posts, and Instagram posts. "How up to date is the data?"