StealAds
Server Details
Search competitor Meta ads, find winning and longest-running ads, and adapt one for your brand.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- TheMattBerman/ads-mcp
- GitHub Stars
- 0
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 58 tools
Most tools have a clearly distinct resource+action, and descriptions actively steer between near-neighbors (e.g. find_long_running_ads vs search_ads, get_competitor_analysis vs list_tracked_rival_ads vs get_playbook, get_ad_status vs get_creative_workspace vs get_campaign_output_workspace). The densest cluster is ad discovery/analysis (search_ads, find_long_running_ads, list_tracked_rival_ads, list_new_competitor_ads, summarize_rival_ad_patterns), where an agent must read carefully to pick correctly. Quote/confirm pairs (quote_steal_ad vs steal_ad, create_campaign preview vs confirm) are distinguishable only because the descriptions spell out the two-step pattern.
Every tool is snake_case with a verb-first convention (list_*, get_*, create_*, update_*, delete_*, start_/pause_/resume_tracking, search_*, save_ad/unsave_ad). Paired operations are symmetrically named (apply/discard/undo_copy_revision, add_competitor, move_saved_ad vs save_ad). The only deviation is whoami, which is a conventional identity call rather than an inconsistency.
58 tools is far above a comfortable surface, even for a fairly broad product spanning brands, competitors, tracking, ad library, generation, copy editing, collections and billing. Many tools are thin variants or read-only siblings of one another (four get_* workspace/status readers, five ad-list/search tools, asset upload split into upload vs register_url). It is scoped to a real domain rather than padded, but the volume itself raises selection cost.
Coverage is broad: full brand lifecycle (create/get/list/update), competitor tracking lifecycle, assets (list/upload/register/update/delete/set primary), campaigns with outputs, version selection and export descriptors, saved ads and collections, copy revision history, usage and capabilities. Gaps are mostly deletes (no delete_brand, delete_collection, delete_competitor, delete_playbook) and no publishing/archiving operation, which agents may hit when housekeeping. These are workaroundable or plausibly out of scope, so minor rather than blocking.
Available Tools
58 toolsadd_competitorAdd a competitorAInspect
Saves an advertiser as a competitor under one of the account's brands and returns it with its competitor id; a matching name or Facebook page returns the existing competitor instead of a duplicate. Adding without track spends no credits. With track: true it also starts tracking. When tracking would charge credits, the call without confirm returns a cost preview (idempotency_key, max_credits) and saves nothing, and the same call with confirm: true and those two values saves the competitor and starts tracking. Tracking that costs no credits starts on the first call. search_advertisers finds the facebook_page_id that pins the right page.
Inputs: name (required), brand_id (required), track, confirm, +5 more.
apply_copy_revisionApply a copy revisionADestructiveIdempotentInspect
Accepts a delivered pending copy proposal from revise_copy and makes it the ad’s current copy. Takes the pending version_id that get_creative_workspace reports. No further charge; undo_copy_revision can bring the earlier copy back.
Inputs: version_id (required).
contact_supportContact StealAds supportAIdempotentInspect
Sends a help request, problem report or feature suggestion from the user to the StealAds team and returns a confirmation that it was saved. For when the user asks for help or reports that something is wrong; it passes the message on and does not answer it. Identical messages within a day are combined, and an account can send up to five new messages per hour.
Inputs: kind (required), message (required).
create_brandCreate a brandAInspect
Creates a brand from the details supplied and returns it with its brand id. kind brand (default) needs name, industry, primaryColor, toneOfVoice and targetAudience; kind research needs only a name and suits competitor research, with details to add before making ads. Does not read the website and spends no credits. The plan's brand allowance applies.
Inputs: name (required), kind, industry, websiteUrl, +3 more.
create_campaignCreate an image campaignAIdempotentInspect
Makes 1 to 10 image ads for one brand in one size. Sources come from sources.type rivals_best (the brand's competitors), source_ad_ids (2 to 5 picks), or automatic selection. product_image_id and prompt set the product image and message. Spends credits only after explicit confirm: true. Two steps: without confirm it returns a cost preview (cost_credits, output_count, source_ad_ids, truth_preflight, confirmation_truth, idempotency_key, max_credits) and spends nothing; the confirming call repeats the same arguments with confirm: true plus the preview's source_ad_ids, count equal to output_count, confirmation_truth as truth, idempotency_key and max_credits. Returns campaign_id and a campaign job reference for get_campaign_outputs. Fewer images than requested can be delivered when sources or quality checks fail. source_preparation_usd estimates source preparation inside the confirmed job; previews spend nothing. Does not publish ads. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation.
Inputs: size, count, truth, prompt, +10 more.
create_collectionCreate a collectionAInspect
Creates a private collection for sorting saved ads and returns it with its collection id. Each call creates a new collection, even when the name repeats. save_ad and move_saved_ad put ads into it.
Inputs: name (required), description.
create_playbookCreate a PlaybookAInspect
Spends credits only after explicit confirm: true. Builds a Playbook for one competitor: a written summary of its stored ads with ideas to try against them. When credits would be charged it takes two steps: without confirm it returns a cost preview (idempotency_key, max_credits) and changes nothing; the same call with confirm: true and those two values starts the analysis and returns a playbook_id. With nothing to charge, the first call starts the analysis, or returns the current Playbook instead of rebuilding it. The analysis runs in the background and get_playbook shows the result. Does not start tracking.
Inputs: competitor_id (required), confirm, max_credits, idempotency_key.
delete_brand_assetDelete a brand assetADestructiveInspect
Removes one image from a brand's stored images for good. Returns deleted, and storage_deleted, which reports whether the underlying file was removed as well; file cleanup can fail separately. Not reversible: the image would have to be uploaded again.
Inputs: asset_id (required).
discard_copy_revisionDiscard a copy revisionADestructiveIdempotentInspect
Rejects a delivered pending copy proposal from revise_copy; the ad’s current copy stays as it is. Takes the pending version_id that get_creative_workspace reports. The charge for the delivered proposal is kept.
Inputs: version_id (required).
find_long_running_adsFind long-running ads in a categoryARead-onlyIdempotentInspect
Samples a category or product type across the query and up to three related angles, returning up to 24 ads (default 12) with captions, hooks, run dates, delivery evidence, strength labels and cautions. For category inspiration; search_ads supports advertiser lookups and paging. One page, no cursor. Ranking uses observed delivery evidence with approximate ordering; matching is broad, and run_duration_days and still_running show each ad’s evidence. No library totals or complete coverage; run time is not proof of sales.
Inputs: query (required), angle, buyer, format, +12 more.
get_adGet an adARead-onlyIdempotentInspect
Returns the details of one ad by id, from this account or the public Ad Library: caption, headline, call to action, media URLs, run dates, delivery evidence, strength label and hook (hook.spoken is a video's opening line, hook.on_screen its opening overlay text). For reading one ad closely before saving or adapting it. Full video transcripts are left out. The detail is JSON in both text and json formats.
Inputs: ad_id (required), analysis_detail, response_format.
get_ad_statusCheck on an ad being madeARead-onlyIdempotentInspect
Returns the progress of an ad being made: state, stage (writing, designing, rendering, checking), whether it is terminal, the finished image and download once ready, or a failure message with any refunded credits. Meant to be read again after poll_after_seconds, until terminal or stop_polling is true; an ad usually takes a few minutes.
Inputs: job_id (required), response_format.
get_automation_statusGet automation statusARead-onlyIdempotentInspect
Returns the Autopilot settings of one brand: whether weekly ad preparation is on, its cadence and day, how many ads each batch may make, and whether the brand has enough source ads for it. Read-only: turns nothing on, changes no setting and starts no batch. feature_enabled false means Autopilot is not switched on in this environment.
Inputs: brand_id (required), response_format.
get_brandGet a brandARead-onlyIdempotentInspect
Returns the full details of one brand this account owns or that is shared with its organization: name, industry, website, colors, tone of voice, target audience, product details, offer and writing rules. Read-only. list_brands finds the brand id. update_brand changes the name, industry, website, primary color, tone of voice and target audience; the other details are edited in the StealAds web app.
Inputs: brand_id (required), response_format.
get_campaign_export_descriptorGet campaign image pack detailsARead-onlyIdempotentInspect
Returns what is needed to assemble a local image pack for one finished campaign image: its ready placements with sizes and image URLs, the ad copy, and naming and tracked-link details. Takes the selected version from get_campaign_output_workspace as expected_version_id. Read-only: downloads, generates and resizes nothing and records no export. The files themselves are saved with the StealAds CLI download-pack command; the URLs returned here are transport links, not a delivered pack.
Inputs: campaign_id (required), generation_job_id (required), expected_version_id (required), response_format.
get_campaign_outputsGet campaign outputsARead-onlyIdempotentInspect
Returns a campaign's status, phase and delivered images (image URL, source ad and aspect ratio for each) for a campaign job reference. Meant to be read again until status is completed or failed. Only delivered images are listed: images that failed generation or the quality check are left out, so results can be fewer than requested, and a failed campaign returns empty results without a reason. Covers campaigns this account owns or that are shared with its organization.
Inputs: campaign_job_id (required), response_format.
get_campaign_output_workspaceReview a campaign outputARead-onlyIdempotentInspect
Returns one finished campaign image with its versions: the available version ids, the selected version, the ad copy and the saved placement sizes. The read that precedes select_campaign_output_version and get_campaign_export_descriptor. Read-only: generates, resizes and exports nothing. Works for completed outputs created by this account.
Inputs: campaign_id (required), generation_job_id (required), response_format.
get_capabilitiesGet available actionsARead-onlyIdempotentInspect
Returns what this connection can do: every available tool with the scope it needs and whether the connection has it, which tools are read-only, page-size caps, request rate, cursor lifetime, the reading allowance, and the rules for cost previews and confirmation. A useful second call after whoami. Reports permissions only; it does not test whether ad generation is currently available.
Inputs: response_format.
get_competitor_analysisAnalyze a competitor’s adsARead-onlyIdempotentInspect
Summarizes the stored ads of one saved competitor: its most common hooks, angles and funnel stages, plus its longest-running ads. For understanding a competitor's overall approach; list_tracked_rival_ads returns the individual ads and get_playbook the written recommendations. Counts cover only the ads StealAds has stored for that competitor, not its complete advertising.
Inputs: competitor_id (required), response_format.
get_creative_workspaceGet creative copy and historyARead-onlyIdempotentInspect
Returns the copy of one finished image ad: the original and current headline, primary text and call to action, any pending rewrite proposal, and up to 25 revision history entries per page with next_cursor. The read that follows revise_copy and precedes applying, discarding, undoing or editing copy. Takes a generation_job_id, not the job reference that get_ad_status takes; list_generation_jobs returns both. Campaign images are read with get_campaign_output_workspace instead.
Inputs: generation_job_id (required), cursor, response_format.
get_playbookGet a PlaybookARead-onlyIdempotentInspect
Returns the latest Playbook for one competitor of one brand: a written summary of that competitor's ads with ideas to try against them. Read-only and reserves no user credits. When none has been created the result is Playbook not found; create_playbook builds one.
Inputs: brand_id (required), competitor_id (required), response_format.
get_tracking_statusGet tracking statusARead-onlyIdempotentInspect
Returns the tracking state of one saved competitor: whether tracking is on, the paid-through date, the next scheduled scan, the last successful scan and the latest scan's status. For checking before start_tracking, pause_tracking or scan_competitor, and for seeing whether a requested scan has finished.
Inputs: competitor_id (required), response_format.
get_usageCheck usage and creditsARead-onlyIdempotentInspect
Returns the account's credits used, included and remaining, when they reset, the plan, a billing link, and the daily reading allowance that saved and tracked-ad reads draw on. For checking the balance before paid work, or after a read reports that the allowance is used up.
Inputs: brand_id, response_format.
list_brand_assetsList brand assetsARead-onlyIdempotentInspect
Lists one brand's stored images, up to 24 per page with next_cursor: each asset id, type (product, lifestyle, model, logo or reference), URL, description, tags and whether it is the primary image of its type. The source of asset_id for the brand image tools and of product_image_id for campaigns. Read-only; uploads and analyzes nothing.
Inputs: brand_id (required), cursor, page_size, asset_type, +1 more.
list_brandsList brandsARead-onlyIdempotentInspect
Lists the brands this account can use, most recently updated first, as short summaries with the brand id that other tools take as brand_id. Returns 25 per page by default (at most 50), with next_cursor when more exist. get_brand returns one brand's full details.
Inputs: cursor, max_results, response_format.
list_campaignsList your campaignsARead-onlyIdempotentInspect
Lists campaigns created by this account with each campaign's id, job reference, brand, name, status and delivered images. The way to find the campaign job reference for get_campaign_outputs and to recover when a response was lost. Returns at most 50 (default 20) as a recent list without a cursor, not full campaign history. Archived campaigns appear only with include_archived: true. Teammates’ campaigns are not listed.
Inputs: brand_id, max_results, response_format, include_archived.
list_collectionsList collectionsARead-onlyIdempotentInspect
Lists the user's saved-ad collections, the private folders that saved ads are sorted into, with each collection id, name, description and ad count, one page at a time with next_cursor. Read-only; list_saved_ads reads the ads inside one.
Inputs: cursor, page_size, response_format.
list_competitorsList competitorsARead-onlyIdempotentInspect
Lists the competitors saved under one brand, up to 50 per page, with each competitor id, Facebook page, tracking state, stored ad count and flags (stale, off_category, possible_duplicate). The source of competitor_id for the tracking, analysis and Playbook tools. Counts cover only the ads stored for this account's tracked set, not an advertiser's full activity; a count still being computed shows as pending with count_retry_at, and off_category false can also mean not yet checked.
Inputs: brand_id (required), limit, cursor, response_format.
list_generation_jobsList your adsARead-onlyIdempotentInspect
Lists the ads made by this account with status, phase and finished image URLs, optionally for one brand. The way to find the job reference for get_ad_status or the generation_job_id for the copy tools, and to recover when a response was lost. Returns up to 50 per page (default 20) with next_cursor; a page can be short or empty while has_more is true. Teammates’ ads are not listed.
Inputs: cursor, brand_id, max_results, response_format.
list_new_competitor_adsNew ads from tracked rivalsARead-onlyIdempotentInspect
Lists the ads StealAds first observed from a brand’s tracked competitors since a date, newest first, up to 50 per page: the tool for what rivals launched this week. since defaults to seven days ago and reaches back at most 30 days. first_observed_at is when StealAds first saw the ad, not its launch date; first_import marks ads from a competitor’s initial import. The final page returns resume_since for the next periodic check. Ads and counts describe only this account’s tracked set, not complete advertiser or library coverage. Each ad returned draws on the daily workspace reading allowance.
Inputs: brand_id (required), since, cursor, format, +4 more.
list_playbooksList PlaybooksARead-onlyIdempotentInspect
Lists the latest Playbook status for each competitor saved under a brand (the first 50 competitors): whether one exists, its status, how many ads it covers and when it was generated. A Playbook is a written summary of a competitor's ads with ideas to try against them. Read-only; get_playbook returns the full text and create_playbook builds one.
Inputs: brand_id (required), response_format.
list_saved_adsList saved adsARead-onlyIdempotentInspect
Lists the user's saved ads, up to 20 per page with next_cursor: each ad with its note, format, run dates and delivery evidence. collection_id reads one collection, null reads the ads not in any collection, and leaving it out reads every saved ad. Saved-ad reads draw on the daily workspace reading allowance.
Inputs: cursor, page_size, collection_id, response_format.
list_tracked_rival_adsList tracked rival adsARead-onlyIdempotentInspect
Pages through the stored ads of the competitors (rivals) a brand tracks, running and stopped, up to 50 per page: caption, opening hook, format, dates and delivery evidence for each ad. The tool for a full walk of tracked competitors, and the only ad list with table (CSV) output. Ads and counts describe only this account’s tracked set, not complete advertiser or library coverage. Order recent (default) mixes all tracked competitors; rival walks one at a time and reports totals.tracked_rivals once the walk completes. Missing hook evidence and unanalyzed why are null. Each ad returned draws on the daily workspace reading allowance. A page can be short or empty while has_more is true; page_note says why.
Inputs: brand_id (required), limit, order, cursor, +6 more.
list_weekly_dropsList weekly dropsARead-onlyIdempotentInspect
Lists a brand's recent Weekly Drops, the weekly batches of ad ideas that Autopilot prepares: up to 12, each with its period, status, result count and timestamps. Read-only; starts no batch. get_automation_status shows the settings behind them.
Inputs: brand_id (required), max_results, response_format.
move_saved_adMove a saved adADestructiveIdempotentInspect
Moves an ad that is already saved into one of the user's collections, or back to unsorted with collection_id null. save_ad saves an ad that is not saved yet. Repeating the same move changes nothing.
Inputs: ad_id (required), collection_id (required).
pause_trackingPause trackingADestructiveIdempotentInspect
Turns off scheduled scans for a tracked competitor and returns tracking: false. Ads already collected are not deleted. The current paid month is not refunded and a scan that is already running is not cancelled. resume_tracking turns tracking back on.
Inputs: competitor_id (required).
quote_steal_adCheck the cost to adapt an adARead-onlyIdempotentInspect
Returns What we’ll make (source, offer, message and all added instructions, with origins) and any OFFER_CONFLICT choice before the cost preview for adapting one source image ad to a brand with steal_ad: the credit cost, current balance, whether the balance covers it, and the idempotency_key for steal_ad's confirming call. Also checks that the source ad and brand can be used. Read-only: creates no ad, reserves no credits and starts no job. source_preparation_usd estimates source preparation inside the confirmed job. A video source gets no quote; image-ad alternatives and a web link for making a Script come back instead. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation.
Inputs: ad_id (required), brand_id (required), notes, language, +3 more.
register_brand_asset_urlImport a brand image URLAIdempotentInspect
Imports an image from a public HTTPS URL into a brand's stored images, up to 10 MiB, and returns the new asset, or a pending status while the import finishes. For images already hosted online; upload_brand_asset takes small files as base64. Does not analyze the image or make it primary. Needs a user sign-in connection, not an API key.
Inputs: brand_id (required), image_url (required), asset_type (required), idempotency_key (required), +2 more.
resume_trackingResume trackingAIdempotentInspect
Turns tracking back on for a saved competitor whose tracking was paused; otherwise the same as start_tracking. When tracking would charge credits it takes two steps: without confirm it returns a cost preview (idempotency_key, max_credits) and changes nothing; the same call with confirm: true and those two values starts tracking. Tracking that costs no credits starts on the first call. A competitor that is already tracked returns tracking: true at no cost.
Inputs: competitor_id (required), confirm, max_credits, idempotency_key.
revise_copyRevise creative copyAIdempotentInspect
Asks StealAds to rewrite the copy of a finished image ad as a proposal; the current copy stays in place until the proposal is applied. Spends credits only after explicit confirm: true. Two steps: without confirm it returns a cost preview (idempotency_key, max_credits) and changes nothing; the same call with confirm: true and those two values starts the rewrite. get_creative_workspace then shows the proposal as pending, and apply_copy_revision or discard_copy_revision settles it. A delivered proposal is charged even when discarded; update_copy saves wording the user supplies without a charge.
Inputs: mode (required), generation_job_id (required), note, confirm, +2 more.
save_adSave an adADestructiveIdempotentInspect
Saves an ad to the user's saved ads (swipe file), optionally into a collection and with a note, and returns the saved entry. Works for ads in this account and public Ad Library ads. Saving an ad that is already saved changes only the collection or note that is supplied and creates no duplicate; it cannot unsort an ad, which move_saved_ad does with collection_id null. list_saved_ads reads saved ads back.
Inputs: ad_id (required), notes, collection_id.
scan_competitorScan a tracked competitorAIdempotentInspect
Requests one immediate check for new ads from a competitor that is already tracked; get_tracking_status shows when it finishes. Two steps: without confirm it returns a preview with idempotency_key; the same call with confirm: true, that key and max_credits: 0 starts the scan. It runs under the existing tracking month and does not start or renew tracking. Refused when tracking is off or paused, when a scan is already queued or running, and within one hour of the last manual scan.
Inputs: competitor_id (required), confirm, max_credits, idempotency_key.
search_adsSearch the Ad LibraryARead-onlyIdempotentInspect
Searches the StealAds Ad Library of public Meta ads by advertiser name, website or keywords, with filters for format, run length, industry, buyer, hook and angle. Returns up to 24 ads per page (default 12) with caption, hook, run dates, delivery evidence and image URLs, plus next_cursor. For a named advertiser or a specific topic; find_long_running_ads suits open category inspiration. Results are a capped sample without library totals or complete advertiser coverage. Keyword matching is broad, so rows can be off-topic. Run time is delivery evidence, not proof of sales.
Inputs: query (required), angle, buyer, cursor, +15 more.
search_advertisersFind public advertisersARead-onlyIdempotentInspect
Looks up public advertisers by name, website domain, Facebook page id or Ad Library URL and returns matching pages with their facebook_page_id, the value add_competitor and the search tools take to pin one advertiser. Confirmed matches come back in advertisers; near misses come back in suggestions and are not confirmed. A brand StealAds has not collected returns status not_indexed with a web link for requesting it; this tool starts no fetch. Results are a capped page with next_cursor.
Inputs: query (required), cursor, max_results, response_format.
select_campaign_output_versionSelect a campaign output versionADestructiveIdempotentInspect
Makes an existing image version the selected one for a campaign image. Takes the target version_id and expected_current_version_id, both from get_campaign_output_workspace; when the selection has changed since that read the call is refused, so a stale choice cannot overwrite a newer one. Selecting the version that is already selected changes nothing. Generates and exports nothing.
Inputs: version_id (required), campaign_id (required), generation_job_id (required), expected_current_version_id (required).
set_primary_brand_assetSet a primary brand imageADestructiveIdempotentInspect
Makes one stored image the primary image of its type for its brand, for example the main product shot, and returns the updated asset. A logo asset also becomes the brand's logo. Repeating the call changes nothing.
Inputs: asset_id (required).
show_ad_galleryShow selected adsARead-onlyIdempotentInspect
Displays up to 12 chosen ads as one visual gallery in clients that render MCP Apps widgets, and returns the same ads' data, image previews and links for text-only clients. For presenting the ads the user should see, once the relevant results are chosen. Each ad id is checked again for access; ads no longer available are left out with incomplete: true. A successful call does not prove that a gallery rendered.
Inputs: ad_ids (required), query, response_format.
start_trackingStart trackingAIdempotentInspect
Turns on tracking for a competitor already saved under a brand, so StealAds collects that advertiser's new ads on a schedule. add_competitor with track: true saves and tracks a new competitor in one call. When tracking would charge credits it takes two steps: without confirm it returns a cost preview (idempotency_key, max_credits) and changes nothing; the same call with confirm: true and those two values starts tracking. Tracking that costs no credits starts on the first call. A competitor that is already tracked returns tracking: true at no cost.
Inputs: competitor_id (required), confirm, max_credits, idempotency_key.
steal_adAdapt an image adAIdempotentInspect
Makes one new image ad for a brand by adapting (remixing) a source image ad. Spends credits only after explicit confirm: true. Two steps: without confirm it returns a cost preview with idempotency_key, as quote_steal_ad does, and spends nothing; the same ad_id, brand_id and notes with confirm: true and that key start the ad and reserve credits. Returns a job reference for get_ad_status; an ad usually takes a few minutes and can fail its quality check, with refunded credits reported there. notes are guidance: the product image is chosen automatically and on-image wording can differ, while create_campaign takes a product image and prompt. Confirmed jobs prepare an unprepared source; source_preparation_usd is its maximum preparation cost. Does not publish ads. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation. Pass language as a BCP-47 tag from the user’s requested output language (including requests in notes/prompt); use regional tags such as es-MX. The quote has no generated copy: never invent proposed copy. Copy is written after approval. Unsupported tags use site language, disclosed in the quote. Echo language unchanged on confirmation.
Inputs: ad_id (required), brand_id (required), notes, confirm, +6 more.
summarize_rival_ad_patternsSummarize rival ad patternsAIdempotentInspect
Summarizes observed ad patterns for one of the account’s brands: up to three brand-specific insights and up to twelve standout image ads (at most three per advertiser), from the brand’s tracked competitors or, with category, from a one-time Ad Library sample that tracks nothing. Patterns describe the sample, not measured sales performance. state insufficient_evidence means fewer than two standout ads were found; a category, or more tracked competitors, widens the evidence. Samples carry no totals and stale observations are marked. Ads returned draw on the daily workspace reading allowance.
Inputs: brand_id, category, response_format.
undo_copy_revisionUndo a copy revisionADestructiveIdempotentInspect
Restores the copy that was current before a kept revision of a finished image ad, as a new history entry, and returns the new version. Takes the version_id of a kept revision from the get_creative_workspace history. Nothing is generated and no credits are spent.
Inputs: version_id (required), idempotency_key (required), generation_job_id (required).
unsave_adUnsave an adADestructiveIdempotentInspect
Removes an ad from the user's saved ads and returns deleted: true; an ad that was not saved returns deleted: false. The ad itself is not deleted and can be saved again with save_ad.
Inputs: ad_id (required).
update_brandUpdate a brandADestructiveIdempotentInspect
Changes one brand's name, industry, website, primary color, tone of voice or target audience and returns the updated brand; fields left out stay as they are, and at least one is needed. Product details, offers, writing rules and other brand settings are not editable here; those are changed in the StealAds web app. Does not read the website, re-run brand discovery or spend credits.
Inputs: brand_id (required), name, industry, websiteUrl, +3 more.
update_brand_assetUpdate brand asset detailsADestructiveIdempotentInspect
Changes the description or tags of one stored brand image and returns the updated asset; at least one of the two is needed. The image file itself is not replaced or analyzed.
Inputs: asset_id (required), tags, description.
update_collectionUpdate a collectionADestructiveIdempotentInspect
Renames one of the user's collections or changes its description; at least one of the two is needed. The saved ads inside it are not affected.
Inputs: collection_id (required), name, description.
update_copyUpdate creative copyADestructiveIdempotentInspect
Saves primary text and/or call-to-action wording supplied by the user for a finished image ad, as a new entry in its copy history. For exact wording the user dictates; revise_copy asks StealAds to write a proposal instead. The headline and the image are not changed, nothing is generated and no credits are spent. At least one of primary_text and cta is needed.
Inputs: idempotency_key (required), generation_job_id (required), cta, primary_text.
upload_brand_assetUpload a small brand imageAIdempotentInspect
Uploads a small PNG, JPEG, WebP or GIF image into a brand's stored images, up to 512 KiB before base64 encoding, and returns the new asset or a pending status. For local files; register_brand_asset_url suits larger images already hosted at a public URL. Does not analyze the image or make it primary. Needs a user sign-in connection, not an API key.
Inputs: brand_id (required), asset_type (required), data_base64 (required), content_type (required), +3 more.
whoamiWho am IARead-onlyIdempotentInspect
Returns the account behind this connection: display name, masked email, granted scopes (read, write, spend), brand count and up to ten brand names. A first call for confirming which StealAds account and permissions are in use; list_brands returns the brand ids that other tools take.
Inputs: response_format.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Search a library of winning Meta ads and remake them in your brand, or make new ones.
Winning Meta ads with Meta's EU reach, scripts and A/B tests. Find a competitor, see what changed.
Create Meta (Facebook and Instagram) ad creatives with AI and research competitor ads.
Search winning Meta ads with $50k+ in estimated spend, read hooks and scripts, and draft briefs.
Related MCP Servers
AlicenseBqualityCmaintenanceSee who is running ads in your space, clone a winning ad for your brand as an image or video, and launch it on Meta, Google, TikTok, X, or ChatGPT Ads. The full research, clone, launch loop from one chat in Claude, ChatGPT, or Cursor. Flat pricing, no fabricated metrics.4730 npmMIT- AlicenseNot gradedqualityDmaintenanceSearches and analyzes competitor ads and content across Meta, Google, Instagram, TikTok, and YouTube with AI-powered creative analysis and cross-platform brand discovery.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables competitor ad research inside AI agents by searching public Meta and TikTok ad libraries, scoring ad copy, and returning active ads with marketing and performance signals.MIT
- AlicenseNot gradedqualityCmaintenancePerplexity for social: ask what competitors are running across social and get answers grounded in real ad and creative data, not guesses.2Apache 2.0
Glama MCP Gateway
Add one secure layer between your agents and this server.