Viral Outliers

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Find viral outlier posts on TikTok, Instagram and YouTube, pull creator stats, and crawl on demand.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "Viral Outliers": {
      "url": "https://viraloutliers.com/api/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (34)

• search_outliers read-only

  Search a continuously-crawled database of social media posts ranked by outlier score: how strongly a post overperforms the account's own baseline. Filter by platform, keyword, exact creator handle, content type, an outlier-score band (min/max), a view band (min/max), a follower band (min/max), an engagement-rate band (min/max) and time window; sort by outlier score, views, likes, engagement or recency. Returns post metadata, stats and thumbnails. Use this to find proven viral formats in any niche before creating content. Results may include deleted posts (deleted_at set) and posts from deactivated profiles (profile is_active=false), these are kept for their thumbnails and format ideas, with stats frozen at deletion; filter on deleted_at / is_active if you only want live content. Cost: 1 credit per 25 posts returned.

  Parameters:

  • page — integer — Page number, 1-based. A search serves its first 1,000 posts; a page past that is refused at no charge, so narrow the filters instead.
  • query — string — Keyword search over captions/titles
  • handle — string — Exact creator handle filter
  • sortBy — string
  • maxViews — number
  • minViews — number
  • pageSize — integer — Results per page, 1 to 100 (default 25). Billed per 25 posts returned: 25 costs 1 credit, 100 costs 4.
  • platforms — array
  • timeFrame — string
  • categoryId — string — Niche filter: a category id from list_categories (covers its subtree)
  • watchlistId — string — Scope to the creators in one of your watchlists (id from list_watchlists)
  • contentTypes — array
  • maxFollowers — number
  • minFollowers — number
  • maxOutlierScore — number
  • minOutlierScore — number
  • maxEngagementRate — number
  • minEngagementRate — number

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• search_profiles read-only

  Search tracked social media profiles by handle or name, filtered by platform. Returns profile metadata, follower counts and average performance stats. Use it to resolve a handle to a profile id before fetching stats or posts, or to discover creators in the database. Results may include deactivated profiles (is_active=false), e.g. an account that was renamed or went private, retained with frozen stats; filter on is_active if you only want live accounts. Note a creator who changed handles can appear as two rows (old deactivated + new active). Cost: 1 credit per 25 profiles returned.

  Parameters:

  • page — integer — Page number, 1-based. A search serves its first 1,000 profiles; a page past that is refused at no charge, so narrow the filters instead.
  • query — string — Handle, @handle, or profile URL
  • pageSize — integer — Results per page, 1 to 100 (default 25). Billed per 25 profiles returned: 25 costs 1 credit, 100 costs 4.
  • platforms — array
  • categoryId — string — Niche filter: a category id from list_categories (covers its subtree); without a query returns that niche's top profiles

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• list_categories read-only

  Returns the curated niche taxonomy (up to four levels, e.g. "Health > Exercise > Gym") with, for each category, its id, path, level, parent and how many active profiles are categorized under it including its sub-niches. Only categories with at least minProfiles profiles in their subtree are returned (default 5), so every id you get back yields a real result set. Pass a categoryId to search_outliers, search_profiles or niche_trends to scope them to the creators in that niche; the filter always covers the subtree. Free to call and cached, so call it at the start of a session and reuse the ids. Categories are assigned per creator by our categorizer (reviewed by hand for new niches), which makes this filter more precise than a keyword over captions: it catches creators whose captions are hashtag-only or not in English. Cost: free.

  Parameters:

  • maxLevel — integer — Only niches up to this depth (default 4)
  • minProfiles — integer — Only niches with at least this many profiles in their subtree (default 5)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• find_related_social_media_profiles

  Discover new creators on TikTok, Instagram or YouTube from one seed: a creator handle (walks who they follow on TikTok, who they collaborate with, tag or mention on Instagram), a hashtag, or a search phrase. Every account surfaced is screened live with one provider call (recent posts: how many reached 5k+ views, posting activity, newest post) and judged for topic relevance from its bio and captions, then returned ranked with recommended = qualified and on-topic. Accounts already in the database come back flagged known with their profileId. Nothing is crawled or added automatically: call crawl_profile for the ones you want. Priced up to 25 credits per call: 5 base plus 1 per account actually screened, the rest refunded (nothing found = fully refunded). Pass topic to say what related means for your use case. Not for looking up a known creator (use search_profiles). Cost: 25 credits per call.

  Parameters:

  • platform (required) — string
  • limit — integer — Candidates to screen and return (default 10, max 20).
  • query — string — Search phrase: creators behind the top results.
  • topic — string — What "related" means for the relevance judge, e.g. "home workouts for women". Defaults to the seed.
  • handle — string — Seed creator handle: walk who they follow (TikTok) or collaborate with / tag (Instagram). Not on YouTube.
  • hashtag — string — Hashtag without #: creators posting under it.

  Hints: read-only no · destructive no · idempotent no · open-world yes

• get_post read-only

  Fetch a single post by id: views, likes, comments, engagement rate, outlier scores for seven time windows, thumbnail and the owning profile. When a transcript or visual analysis already exists it is included at no extra cost. The visual analysis is a structured scene-by-scene breakdown (per-scene timing, on-screen text, visual elements and a recreation note) plus an overall-style summary. Request new enrichment via request_transcript (speech / on-screen text) or request_visual_analysis (scene breakdown). Use after search_outliers to deep-dive a result. Cost: 1 credit per call.

  Parameters:

  • postId (required) — string — Internal post id from search results
  • includeTranscript — boolean — Attach cached transcript (default true)
  • includeVisualAnalysis — boolean — Attach cached visual analysis (default true)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_profile read-only

  Fetch one tracked profile by id: follower count, bio, average views/likes/engagement across time windows, and recent tracked posts. The averages are the baseline outlier scores are computed against. Use search_profiles first to resolve a handle to an id. Cost: 1 credit per call.

  Parameters:

  • handle — string
  • platform — string — With handle: look up by handle instead of id
  • profileId — string — Profile id from search results

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• request_transcript

  Queues AI transcription of the spoken audio for a video post. Accepts a public post URL or an internal post id. Asynchronous: returns a job reference to poll with get_job_status; once complete the transcript is attached to get_post responses. Credits are charged on queueing and automatically refunded if the job fails. Image slideshows and photo posts have no audio and are rejected up front with no charge: use request_visual_analysis for their on-screen text instead. Cost: 10 credits per call.

  Parameters:

  • url — string — Public post URL (alternative to postId)
  • postId — string — Internal post id

  Hints: read-only no · destructive no · idempotent no · open-world yes

• request_visual_analysis

  Queues a structured visual analysis for a post: for a video, a scene-by-scene breakdown (per-scene timing, scene type, on-screen text, visual elements and a recreation note) plus an overall-style summary (color palette, text style, editing pace); for an image slideshow, per-slide text and visual descriptions. Accepts a public post URL or an internal post id. Asynchronous: returns a job reference to poll with get_job_status; once complete the analysis is attached to get_post responses (includeVisualAnalysis). Credits are charged on queueing and automatically refunded if the job fails. This is the visual twin of request_transcript. Cost: 10 credits per call.

  Parameters:

  • url — string — Public post URL (alternative to postId)
  • postId — string — Internal post id

  Hints: read-only no · destructive no · idempotent no · open-world yes

• crawl_profile

  Submits a public profile for crawling: profile metadata plus its recent posts, stats and thumbnails, after which it stays tracked and appears in searches. Asynchronous. Returns a job reference for get_job_status. If the profile is already tracked, this returns immediately without charging a full crawl. Credits are refunded automatically when a crawl fails. Cost: 40 credits per call.

  Parameters:

  • platform (required) — string
  • handle (required) — string

  Hints: read-only no · destructive no · idempotent no · open-world yes

• get_job_status read-only

  Returns the status of an asynchronous job started by crawl_profile or request_transcript: pending, processing, completed or failed. Free to call: polling must never cost credits. Poll every 10–30 seconds; jobs typically complete within a few minutes. Cost: free.

  Parameters:

  • jobRef (required) — string

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_credit_balance read-only

  Returns the current API credit balance for the authenticated account, plus the automatic top-up state of the account (autoTopup: configured, enabled, pack, monthly charge cap and count, last error) so an agent can tell whether a zero balance is a hard stop or refills itself on the next billable call. Free to call. Agents should check the balance before starting large batch jobs and surface "insufficient_credits" errors to the user with a link to top up. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• compare_profiles read-only

  Compares 2–5 tracked profiles (by profile id or platform+handle) and returns each one's stats (follower count, average views/likes across time windows, engagement rate) plus a computed ranking flagging the top performer by followers and by engagement. Pure lookup, fast. Unresolved profiles come back in a notFound list rather than failing the call. Cost: 2 credits per call.

  Parameters:

  • handles — array — Handle pairs to resolve; 2–5 profiles total across both fields
  • profileIds — array — Profile ids to compare

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• niche_trends read-only

  Returns the top overperforming posts in a niche over a recent window (default one week), plus a breakdown of which content types are driving the trend and which creators are represented. Filter by keyword and platform. Great as the first call in a content-research loop: see what's hot, then deep-dive or remix the winners. Cost: 2 credits per call.

  Parameters:

  • limit — integer
  • query — string — Niche keyword
  • platforms — array
  • timeFrame — string — Default one_week
  • categoryId — string — Niche filter: a category id from list_categories (covers its subtree)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• resolve_post_url read-only

  Takes a public post URL (TikTok video/photo, Instagram post/reel, YouTube video/short) and resolves it to the tracked post: the entry point when your input is a link. found:true returns the postId for get_post, request_transcript, download_post_media and remix_post. found:false tells you whether the whole profile is untracked (call crawl_profile) or just this post. Note that remix_post accepts an untracked URL directly and ingests it itself, so for a remix you can skip this lookup. TikTok/Instagram short links (vm.tiktok.com, /share/) must be expanded to the canonical URL first. Cost: 1 credit per call.

  Parameters:

  • url (required) — string — Public post URL (canonical, not a short link)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• download_post_media

  Returns direct media URLs for a tracked post, the video file, or slideshow images with positions. Accepts a public post URL or an internal post id. When media is not stored yet it queues an on-demand fetch and returns a jobRef to poll (typically ready within ~90 seconds); call again once complete. Platform CDN URLs can expire, so download promptly. YouTube currently returns the thumbnail image. You are responsible for using downloaded media in line with the platforms' terms and applicable law. Cost: 3 credits per call.

  Parameters:

  • url — string — Public post URL (alternative to postId)
  • postId — string — Internal post id

  Hints: read-only no · destructive no · idempotent no · open-world yes

• remix_post

  Takes a post (pass a public post URL or an internal post id) plus a target niche/brand description, and produces an adapted content idea: rewritten title and description, segment-by-segment script, a short analysis of why the original went viral, and an execution checklist, the proven format transplanted into your niche. The URL does NOT need to be tracked: an unknown post is fetched and ingested automatically as part of the job, so any public TikTok/Instagram/YouTube link works with no crawl_profile round-trip. Asynchronous (~1–3 minutes, a few minutes longer for an untracked URL): returns a jobRef; poll get_remix_result until its "remix" field is populated (it returns status + result together). For video posts the remix auto-generates the transcript and visual analysis first, so quality no longer depends on you transcribing beforehand. Credits are refunded automatically if the remix fails. Cost: 20 credits per call.

  Parameters:

  • targetNiche (required) — string — Your niche/brand/audience, 5–500 chars
  • url — string — Public post URL (alternative to postId — what a user browsing social media has)
  • tone — string — Preferred tone of voice
  • postId — string — Internal post id

  Hints: read-only no · destructive no · idempotent no · open-world yes

• get_remix_result read-only

  Returns the finished remix for a remix_post job: adapted title, description, script segments, why-it-went-viral analysis (whyItWorks) and checklist. Free to call. Poll get_job_status until the job reports completed, then fetch here. The result is also viewable in the web app under Content Ideas. Cost: free.

  Parameters:

  • jobRef (required) — string — The pipeline jobRef returned by remix_post

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• create_topup_link

  Creates a Stripe Checkout link for a credit pack, so when the balance runs out mid-task you can hand the account owner a one-click payment link instead of instructions. Credits land within seconds of payment. Free to call (5 links/hour); links are valid for 24 hours. Packs: see the pricing table in the docs. Cost: free.

  Parameters:

  • packId (required) — string — Credit pack to buy — see get_credit_balance / docs for sizes

  Hints: read-only no · destructive no · idempotent no · open-world yes

• report_issue

  Report a problem with any skill: an error you keep hitting, data that looks wrong or stale, or something you needed that the API could not do. Free: never spend credits on telling us something is broken. Include what you called, what you expected and what happened. Set wantsUpdate to true to get an email on your account address when the issue is resolved. Reports go straight to the team. Cost: free.

  Parameters:

  • message (required) — string — What you called, what you expected, what happened (10–2000 chars)
  • skill — string — Which skill the issue is about, when known
  • wantsUpdate — boolean — true = email the account owner when resolved

  Hints: read-only no · destructive no · idempotent no · open-world no

• track_profile

  Starts monitoring a profile so it is automatically re-crawled at a cadence you choose (daily, every_3_days or weekly), keeping its posts and stats fresh without you polling crawl_profile. Identify the profile by profileId, platform+handle, or a public profile/post URL. Managing monitoring is free; each scheduled refresh crawl costs credits (10 per refresh) and monitoring pauses itself if your balance runs out, then resumes when you top up. The profile must already be in the database; crawl_profile it first if it is not. Calling again on an already-monitored profile just updates the cadence. Pull the new posts with get_tracked_updates. Cost: free.

  Parameters:

  • url — string — Public profile or post URL identifying the profile
  • handle — string — Public handle, without @ (pair with platform)
  • platform — string — With handle: resolve the profile to monitor
  • frequency — string — Refresh cadence (default weekly)
  • profileId — string — Internal profile id from search results

  Hints: read-only no · destructive no · idempotent no · open-world yes

• untrack_profile

  Stops monitoring a profile: no further scheduled refresh crawls are charged for it. Free to call. Pass the internal profileId (from list_tracked_profiles or search results). The profile and its already-crawled posts stay in the database and searchable; only the recurring refresh stops. Returns not_found if the profile is not currently being monitored. Cost: free.

  Parameters:

  • profileId (required) — string — Internal profile id to stop monitoring

  Hints: read-only no · destructive no · idempotent no · open-world no

• list_tracked_profiles read-only

  Returns every profile you are currently monitoring, with its handle, platform, follower count, refresh cadence, next scheduled crawl time, when it was last checked for updates, and any paused reason (e.g. paused for insufficient credits). Free to call. Use it to audit what is being refreshed and what it is costing, or to grab a profileId for untrack_profile. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• create_viral_alert

  Creates a viral alert: we watch a TikTok, Instagram or YouTube profile and email the address you give when it publishes a post whose outlier score (views divided by the profile's own average) reaches minOutlierScore (default 3, minimum 1.5) within postAgeDays of posting (default 7, max 30). Identify the profile by profileId, platform+handle, or a public URL; it must already be in the database (crawl_profile it first if not). Setting up, listing and deleting alerts is free. Two things cost credits: the refresh crawl that keeps the profile fresh (10 credits per run, daily by default, started for you if you are not already monitoring the profile with track_profile) and each alert email actually sent (5 credits). Limits: one alert email per destination address per day, sent around sendHour in your timezone (defaults 9, UTC) and bundling every alert that fired; 3 emails per alert in total, after which the alert ends; expiresInDays (default 30, max 90) ends it regardless; 25 active alerts per account. A post can trigger an alert only once. Your account email needs no confirmation; any other address receives a confirmation link first and nothing is sent until it is clicked. Every alert email has its own stop link. Calling again for the same profile and address updates the settings instead of creating a duplicate. Cost: free.

  Parameters:

  • email (required) — string — Where to send the alert. Your account email needs no confirmation; any other address gets a confirmation link first
  • url — string — Public profile or post URL identifying the profile
  • handle — string — Public handle, without @ (pair with platform)
  • platform — string — With handle: resolve the profile to watch
  • sendHour — integer — Preferred hour of day for the email, 0-23 (default 9)
  • timezone — string — IANA timezone for sendHour, e.g. Europe/Brussels (default UTC)
  • frequency — string — Refresh cadence for the profile if tracking is started by this alert (default daily)
  • profileId — string — Internal profile id from search results
  • postAgeDays — integer — Only posts published within this many days can trigger (default 7, max 30)
  • scoreWindow — string — Which profile average the score is measured against (default three_months)
  • expiresInDays — integer — The alert ends after this many days (default 30, max 90)
  • minOutlierScore — number — Trigger at this outlier score or above (views / the profile average; default 3, minimum 1.5)

  Hints: read-only no · destructive no · idempotent no · open-world yes

• list_viral_alerts read-only

  Returns your viral alerts (active first, then recently ended, up to 100) with the watched profile, destination email and whether it is confirmed, threshold and score window, post age window, preferred send hour and timezone, emails sent out of the cap, paused or ended reason, expiry and the last email time. Free to call. Use it to audit what is being watched, to check whether a third-party address has confirmed, or to grab an alertId for delete_viral_alert. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• delete_viral_alert

  Ends a viral alert by alertId (from list_viral_alerts or create_viral_alert). Free. If this alert is what started the profile's tracking and no other active alert of yours needs it, the paid refresh crawls stop as well; tracking you started yourself with track_profile is left untouched. Returns not_found when the alert does not exist or already ended. Cost: free.

  Parameters:

  • alertId (required) — string — Alert id from list_viral_alerts or create_viral_alert

  Hints: read-only no · destructive no · idempotent no · open-world no

• create_watchlist

  Creates an empty watchlist (a named set of creators) and returns its watchlistId. Fill it with add_watchlist_profiles, then pass the id to search_outliers as watchlistId to scope any search to exactly those accounts. Watchlists created here appear in the owner's web app too. Free to call. Watchlists count against a workspace allowance that comes from a subscription or is earned from API spend (every API key holder starts with 1); an over-limit call explains what unlocks more. Cost: free.

  Parameters:

  • name (required) — string — Display name, 1-80 characters
  • notes — string — Optional notes, up to 500 characters

  Hints: read-only no · destructive no · idempotent no · open-world no

• list_watchlists read-only

  Returns every watchlist on the account (both agent-created and web-app-created) with its watchlistId, name, notes, profile count and last-updated time. Free to call. Use it to recall an id for search_outliers, add_watchlist_profiles or get_watchlist. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_watchlist read-only

  Returns a single watchlist by watchlistId with every member profile: profileId, handle, platform and follower count. Free to call. Use it to review or explain a list before searching it, or to pick profileIds for remove_watchlist_profiles. Cost: free.

  Parameters:

  • watchlistId (required) — string — Watchlist id

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• add_watchlist_profiles

  Adds creators to one of your watchlists. Pass profileIds (from search_profiles or any search result) and/or handles as {platform, handle} pairs; up to 25 per call. Creators must already be tracked (crawl_profile first if not; unresolved ones come back in notFound). Already-present creators are reported, not duplicated. Free to call. Each added creator counts against your followed-profiles allowance, which comes from a subscription or is earned from API spend; an over-limit call explains what unlocks more. Cost: free.

  Parameters:

  • watchlistId (required) — string — Target watchlist id
  • handles — array — Handle pairs to resolve and add (profileIds + handles ≤ 25 per call)
  • profileIds — array — Internal profile ids to add

  Hints: read-only no · destructive no · idempotent no · open-world no

• remove_watchlist_profiles

  Removes the given profileIds from a watchlist and returns how many were removed plus the new profile count. Free to call. Use get_watchlist to find the profileIds. Cost: free.

  Parameters:

  • watchlistId (required) — string — Watchlist id
  • profileIds (required) — array — Internal profile ids to remove

  Hints: read-only no · destructive no · idempotent no · open-world no

• delete_watchlist

  Deletes one of your watchlists including its memberships, and frees the watchlist slot and followed-profile slots it used. Free to call. Works for lists created through the API or in the web app; only the owner can delete a list. Cost: free.

  Parameters:

  • watchlistId (required) — string — API-created watchlist id

  Hints: read-only no · destructive no · idempotent no · open-world no

• get_tracked_updates read-only

  Returns the posts newly discovered (first stored by our crawler) since your last check, across all the profiles you are monitoring. Free to call. Each call advances a per-profile cursor, so a subsequent call only returns posts crawled after it (a feed, not a re-scan). New posts arrive when a profile's scheduled refresh crawl runs, on the cadence set by track_profile. Results are the same flattened post shape as search_outliers (stats, thumbnail, handle). Total posts are capped (limit 1–100, default 50). Cost: free.

  Parameters:

  • limit — integer — Max posts to return (default 50)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_trending_outliers read-only

  Returns the current top trending outlier posts across TikTok, Instagram and YouTube, deduplicated to one per creator. Free and unauthenticated (rate-limited per IP; cached ~2 hours). A taste of the database. For filtered search, transcripts and on-demand crawling, create an API key. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_pricing read-only

  Returns the current price list as structured JSON: the USD value of one credit, every skill with its credit cost and REST route, the one-time credit packs, and every subscription plan with its monthly and yearly price, included API credits per month, and watchlist limits. Free and unauthenticated, so an integrator can render live costs to their own users, or an agent can compare subscribing against buying packs, without hardcoding prices or holding an API key. Billable calls also return X-Credits-Charged and X-Credits-Balance response headers, so a reseller can attribute spend per call. Cost: free.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

Publisher website

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.