Evertune

The full verbatim text of one AI response, by the response_id a example_responses row carried. Long responses are hard-truncated. Returns: the response's metadata followed by its full text

response_content

Run one GAQL query against one of this organization's connected Google Ads accounts and return the rows as JSON, as the Google Ads API defines them. Read-only. A query repeated within an hour is answered from the cache. Use google_ads_fields first for resource and field names. Requires Ads Manager access; scoped to the accounts the caller's grant covers.

google_ads_query

The Google Ads API's queryable resources, or one resource's fields (data type, whether each field can be selected, filtered and sorted, enum values, and which resources it can be selected with). Use it to write a valid GAQL query for google_ads_query. Requires Ads Manager access.

google_ads_fields

Keyword-sentiment scores of a custom survey tracker — how the keywords extracted from the models' ANSWERS behave: per keyword, mentions, visibility_score and share_of_answers (0-100), the overall sentiment label and signed sentiment_score (-100..100), with deltas vs the previous report, most-visible first (up to 100 keywords). A different metric from word association (word_association_keywords_for_tracker), which scores how strongly each word is associated with a BRAND. Empty for trackers that have no custom survey behind them. Returns: table of the tracker's keywords with their sentiment metrics and deltas

keyword_sentiment_for_tracker

List the brands you can access in your organization. Call this first to discover the exact brand name to pass to the other tools.

list_brands

The company's published Intercom help-center articles, most recently updated first, one page at a time. Follow next_page until it is null to see every article.

list_intercom_articles

Call one read operation of the OpenAI Ads API for one of this organization's connected ad accounts and return its JSON response verbatim. A call repeated within an hour is answered from the cache. Use openai_ads_api_operations first for the operation ids and parameters; pass array query parameters as repeated name=value entries. Requires Ads Manager access; scoped to the accounts the caller's grant covers.

openai_ads_api_call

Every read operation of the OpenAI Ads API, from OpenAI's published API description (refreshed daily): operation id, method and path, what it returns, and its path and query parameters. Use it to pick an operation and its parameters for openai_ads_api_call. Path parameters (campaign, ad group, ad, audience, feed ids) come from the ids in list results. Requires Ads Manager access.

openai_ads_api_operations

Return daily OpenAI Ads performance (impressions, clicks, spend, conversions, CTR, CPC, CPM) for this org, one row per day and entity, optionally split by device, country, platform or product, with every other field OpenAI offers for that level (status, budgets, names of the parents, ad title/copy/link, product details...) as extra columns. Use to analyse trends and identify high/low performing days. For totals over a window or per month, use ads_campaign_metrics_aggregated: it sums these days for you. Scoped to the ad accounts the caller's ads_access grant covers: a SPECIFIC grant returns only those accounts, not the whole org, so totals can be a subset of the organization's spend — each row's account_id says which account it belongs to. An ALL grant (or admin role) sees every account. Provide start_date/end_date (YYYY-MM-DD) for a specific window, or use lookback_days for a rolling window relative to today. end_date is exclusive. Results are the MOST RECENT rows first (by date, then entity) and are capped; the header shows "total_records: N (showing M)". When N > M the result is truncated — narrow it with a shorter window or specific campaign_ids / ad_group_ids / ad_ids to see the rows you need, rather than trusting the partial set for whole-account comparisons. Spend is in the account currency (shown per row). ctr/cpc/cpm are recomputed from each row's totals. Conversions are not reported by platform or product; those rows show n/a. Returns: table of date, name, account_id, currency, entity_id, campaign_id, ad_group_id, segment (when split), impressions, clicks, spend, conversions, ctr, cpc, cpm, then one column per extra field — one row per (day, account, entity, segment value).

ads_campaign_metrics

Return OpenAI Ads performance TOTALS over any window: one row per entity (and segment value, when split by device, country, platform or product) for the whole window or per calendar month, with impressions, clicks, spend and conversions summed over the days and CTR, CPC, CPM recomputed from those sums. Every other field OpenAI offers (status, budgets, names, ad copy, product details...) shows each distinct value seen in the period with the number of days it held, most days first: "active (20) | paused (11)". A true/false field also gets any_<field> and all_<field> columns, and keeps its counts only when its days disagree. Use for "how did each campaign do last month", comparisons between entities, and any question that needs a sum — no arithmetic left to do. For day-by-day trends use ads_campaign_metrics. Same scope, window and filters as ads_campaign_metrics: the caller's ads_access grant, start_date/end_date (YYYY-MM-DD, end exclusive) or lookback_days, and optional campaign_ids / ad_group_ids / ad_ids. Rows are ordered by month (newest first) and then by spend, largest first, and capped; the header shows "total_records: N (showing M)". Conversions are n/a when no day reported any (always for platform and product). Returns: table of date (the month, per=month only), name, account_id, currency, entity_id, campaign_id, ad_group_id, segment (when split), impressions, clicks, spend, conversions, ctr, cpc, cpm, then one column per extra field.

ads_campaign_metrics_aggregated

Prompt-level brand metrics and previous-report deltas within one CUSTOM tracker topic. Returns AI Brand Score, Visibility Score and Share of Answer on 0-100, Average Position, and Mention Count. A negative Average Position delta means the rank improved. Industry trackers return an explicit unsupported result because their scores have no prompt grain. Returns: bounded prompt-brand rows plus exact pre-limit totals

prompt_scores_for_topic

List pages users saved for this brand, including URLs absent from citation reports. Favorites express preference, not citation performance or ownership. Missing citation rows do not establish zero citations. Read page content before making claims about it.

favorite_urls_for_brand

Search the company's published Intercom help-center articles for a phrase. Follow next_page until it is null to see every match.

search_intercom_articles

PER-RETAILER price breakdown for a product from the AI models' shopping results, most-mentioned retailer first, from each tracker's latest shopping report. Use this when the question is how a product's price differs BY retailer — shopping_products only returns one overall price range per product. Returns: table of (product, retailer, mentions, min/max/avg/median price, currency) rows

shopping_product_prices

Products the AI models recommended in shopping results for the brand's prompts, most-mentioned first, from each tracker's latest shopping report. Per product: mentions (deduped), visibility score (share of shopping- triggered responses that surfaced it — None under a models filter, whose denominator is unknowable), mention share (share of all product mentions), min/max/average/median price in the product's majority currency, average review rating and the retailers carrying it. Scoped by the brand's prompts, not by product owner — competitors' products recommended in those answers are included; filter by `search` for a specific product line. Data source note: calls WITHOUT topic/prompt/retailers filters read the dashboard's saved rollups (complete history, fast); filtered calls read the granular rows, which only cover reports persisted with response ids — totals can differ slightly between the two shapes for trackers whose newest data predates 2026-05-28. Returns: scope totals plus a table of products with shopping metrics

shopping_products

How often the AI models produced ANY shopping results (product carousels, for any brand's products, not this brand's alone) for the brand's prompts — the shopping trigger rate by tracker and topic, from each tracker's latest shopping report. This is also the response denominator behind every shopping visibility score. Returns: table of (tracker, topic) rows with total responses, shopping responses and the trigger rate

shopping_frequency

Purchase links (retailer URLs) from the AI models' shopping results for the brand's prompts, most-mentioned first, from each tracker's latest shopping report. Returns: table of (product, retailer, buy link, mentions) rows

shopping_retailer_urls

Retailers the AI models pointed shoppers to for the brand's prompts, most-mentioned first, from each tracker's latest shopping report. Per retailer: mentions (deduped), visibility score and mention share. Scoped by the brand's prompts, not by product owner — retailers carrying competitors' recommended products are included too. Returns: scope totals plus a table of retailers with shopping metrics

shopping_retailers

First/last-seen ages for specific domains or URLs, computed across the brand's FULL report history — the age columns split out of the source-listing tools. First list sources with domains_for_brand / urls_for_brand (or the tracker/topic variants), then pass the ones you care about here. Each row: first_seen_days_ago (days since the source first appeared in ANY of the brand's reports) and last_seen_days_ago (0 = a report generated today). A source with no sighting in the report history is omitted. NOTE: this scans the whole report history, so keep `sources` to the handful you actually need rather than aging an entire listing. At most 100 distinct sources per call — an oversized request is REJECTED (never truncated), so split larger sets into batches. Returns: a table of sources with first_seen_days_ago / last_seen_days_ago

source_seen_ages

Scores on one topic for every brand scored on it (averaged across trackers when the topic recurs), best first: AI Brand Score, Visibility Score, Average Position and Share of Answer (custom surveys only — null elsewhere), with deltas. Returns: table of brands with their scores and deltas on the topic

topic_scores_vs_competitors

Per-topic scores for a tracker's owner brand. Topics span the AI Brand Index, consumer-preference attributes, custom-survey topics and word-cloud associations. Each topic carries AI Brand Score, Visibility Score, Average Position and Share of Answer (custom surveys only — null elsewhere), with deltas. Returns: table of topics with scores and deltas

topics_for_tracker

Tracker-level scores for every brand scored in a tracker (owner plus competitors), best first: AI Brand Score, Visibility Score, Average Position and Share of Answer (custom surveys only — null elsewhere), with deltas. Returns: table of brands with their scores and deltas on the tracker

tracker_scores_vs_competitors

Word-association keywords for the brand — the words AI models associate with it — aggregated across its trackers and word cloud surveys. Sentiment is signed (negative = unfavorable); association_score is 0-100; deltas are vs the previous run. This is NOT keyword sentiment (see keyword_sentiment_for_tracker): these rows score how strongly a word is associated with a BRAND. Word associations exist per brand (and per tracker) ONLY — there is no topic-level cut anywhere in the product. For a topic-scoped question use the topic score or keyword sentiment tools, and tell the user explicitly that topic-level word association does not exist; never present these brand-level words as if they were scoped to a topic. Returns: table of keywords with sentiment and association scores

word_association_keywords_for_brand

Word-association keywords of a tracker — the words AI models associate with each brand scored in it (owner and competitors): keyword, signed sentiment and 0-100 association score, sorted by brand then descending association, with deltas vs the previous run. Works for both kinds of tracker that produce word associations — a product category tracker's word cloud and a word cloud survey — so just pass the tracker name. For a custom survey's keyword-sentiment scores (mentions, visibility, share of answers) use keyword_sentiment_for_tracker instead: that is a different metric, not a different tracker kind. Word associations exist per brand within a tracker ONLY — there is no topic-level cut anywhere in the product. For a topic-scoped question use the topic score or keyword sentiment tools, and tell the user explicitly that topic-level word association does not exist; never present these brand-level words as if they were scoped to a topic. Returns: table of the tracker's word-association keywords per brand

word_association_keywords_for_tracker