accounts_search
ChatGPTType-ahead search for accounts by name, scoped to the caller's organization and to a project the caller is permitted to see. Backed by lib/accounts/account-search.ts -- the same search the account re-target picker uses, not MCP-only data. Merged-away accounts are excluded. Returns id and name only; no other account fields.
ads_campaign_performance
ChatGPTAnswers "which ad campaigns are working?" for one product over the last windowDays days (7, 28 or 90; default 7). Google Ads: up to 25 campaigns by spend, each with its latest status, spend, impressions, clicks, CTR (0-1), cost per click, conversions and cost per conversion, plus Google totals and when Google last synced. Meta Ads, LinkedIn Ads and Reddit Ads: channel totals only, because their stored rows carry no campaign names. Money is in major units beside its currency; a ratio with nothing under it (CTR with no impressions, cost per conversion with no conversions) is null, and a channel with no stored rows is not_reported with null figures, never 0. Reads only what Runner's daily syncs stored; never calls an ad platform. Connection state, budgets and yesterday's spend are ads.overview; plans awaiting approval are ads.pending_changes. Read-only.
ads_library_read
ChatGPTReads what an advertiser or keyword shows in a public ad library, through Runner's one implementation. library is meta or linkedin; query is a keyword or an advertiser's name. meta reads Meta's official Ad Library API with the product's connected Meta account and answers rows (advertiser, format, hook, cta, firstSeen, stillRunning, url). linkedin is never fetched — its terms do not allow automated reading — so it answers status refused with reason terms_forbid_automated_read, the library's own searchUrl for a person to open, and personReads: the dated reads people on this product already recorded. Other reasons: no_meta_connection, meta_access_review_pending (Meta has not granted this account the archive yet), failed. Every answer carries searchUrl and personReads, so there is always a next step.
ads_list_campaigns
ChatGPTLists the campaigns, ad sets (ad groups) and, when asked, ads of one ad channel of product projectId, each with the exact id ads.draft_budget_change (campaignId) and ads.draft_pause (resourceId, with level) take. channel: any ad channel. includeAdSets (default true) adds the level below campaigns; includeAds (default false) adds ads; campaignId narrows to one campaign and what sits under it; limit 1-100 (default 50), and truncated says more exist. Per entity: id, level (campaign, ad_group or ad: ads.draft_pause's own level; ad_group is Meta's ad set, LinkedIn's campaign, Vibe's strategy) with the platform's own levelLabel, parentId, name, status (active, paused, removed, draft, other, unknown) beside the platform's own platformStatus word, and holdsDailyBudget with dailyBudget and currency: the id whose holdsDailyBudget is true is the one ads.draft_budget_change takes. readiness says whether an ad account is chosen (connected), live changes are on (writesOn), the channel is turned off for the product, and whether the drafting tools can file a plan for this caller now (draftReady, else notReadyReason). source: live is the platform read now (every channel but Google, when its account is ready); synced is Google's daily sync (Google budgets read live for owners and admins; Google ads aren't listed); reported is ids from the daily spend rows when the live read can't run, with no name, level or status, never guessed. Read-only: it never changes or spends anything. Anyone who sees the product's ads in Runner.
ads_overview
ChatGPTAnswers "are our ads running, and what are we spending?" for one product across Google Ads, Meta Ads (Facebook and Instagram), LinkedIn Ads and Reddit Ads. Per channel: whether it is connected, the ad account's name and currency, whether Runner may change it (writesEnabled), active campaigns, yesterday's reported spend, change plans awaiting approval, and totals over the last windowDays days (7, 28 or 90; default 7): spend, impressions, clicks, CTR (0-1), cost per click, conversions and cost per conversion. Also the product's managed ad budget: amount, spent to date, remaining, today's cap, and paceVsEven (spend to date against an even spread of the budget across its dates, within 10% reads even). Money is in major units beside its currency; a figure the platform has not reported is null with a window state of not_reported, never 0. Reads only what Runner's daily syncs stored; never calls an ad platform. Per-campaign detail is ads.campaign_performance; plans awaiting approval are ads.pending_changes. Read-only: it changes nothing.
ads_pending_changes
ChatGPTAnswers "what ad changes are waiting for our approval?" for one product: drafted change plans on Google Ads, Meta Ads, LinkedIn Ads and Reddit Ads that have not expired, newest first (up to 25). Each plan lists its operations (for example "create Campaign", "update Ad set"), how much it moves the daily budget (netDailyBudgetDelta, major units beside currency; null currency means the account's currency is not stored), approvals so far against approvals required, and when it expires. This tool never approves, rejects or applies anything: approving happens in Runner, on the product's Ads page (approveIn, path /ads), behind the approver's second factor. Does not cover spend or performance (ads.overview, ads.campaign_performance). Read-only.
ai_usage_quote
ChatGPTSays what calls would cost before you make them, and spends and changes nothing. items lists the calls you plan, each a tool (media.transcribe, growth.brief_posts, growth.brief_article or growth_team.share_document, dotted or with an underscore) and a quantity (default 1). For each item: quotable; worstCasePerCall and worstCase (quantity times it) in USD — the most the call is held for on the product's AI meter before the vendor is asked, at today's price, so the call costs that or less; paidBy (product_key: the product's own key; runner_credits: Runner's AI credits; runner: Runner's own key); budget, the AI budget that binds it this period (limit, committed including calls in flight, remaining), null when none is set; creditsAvailable on Runner's credits; sufficientBudget; and reason with message when there is no figure (no_key, blocked, unpriced, not_quotable, unreadable). totalWorstCase and sufficientBudget at the top cover every item together. Then call the tool with approvedMaxUsd set to the figure the person approved (at least worstCasePerCall): it refuses over_approved_max rather than spend past it. Every other Runner tool spends nothing from the AI budget. The projectId comes from projects.list.
analytics_search_queries
ChatGPTAnswers "what are people searching for when they find us on Google?" for one product: the top 25 Google Search Console queries by clicks over the last 28 days (Search Console's own recent window; fixed), each with clicks, impressions, click-through rate (ctr, 0-1) and average position (impression-weighted), plus totals and the sync's freshness. Reads only what Runner's daily Search Console sync stored. state says why a list is empty: not_connected (no Search Console property connected), not_synced_yet (connected, nothing stored yet), unavailable (the stored rows could not be read) or checked (read; an empty list then means no searches were recorded). truncated means the busiest 2,000 daily rows were read and totals are floors. Search Console itself leaves out rare queries for privacy. Does not cover site traffic (analytics.traffic_summary) or paid search (ads.campaign_performance). Read-only.
analytics_traffic_summary
ChatGPTAnswers "how's traffic?" for one product's website over the last windowDays days (7, 28 or 90; default 7): visits counted by Runner's own consent-gated collector (total, the same-length period before when the read covers it, and per day); the top traffic sources (campaign tags, UTM source/medium, referring sites, or direct), counted in human page events with bots and unclassified traffic excluded; and, when a GA4 property is connected, the sessions and active users GA4 reported per day as stored by Runner's daily sync. GA4 state is not_connected, not_synced_yet, connected or unavailable — a GA4 figure that was never stored is null, never 0. Visit history covers 56 days, so a 90-day request reports 56 days of visits and says so in notes. Does not identify visitors, does not cover search queries (analytics.search_queries) or ad spend (ads.overview), and never calls Google live. Read-only.
approvals_list_waiting
ChatGPTLists what waits for a decision on product projectId, soonest to expire first (up to 50): proposalId, versionNumber and checksum (the exact version waiting), kind and kindTitle (the kind of action), description (one line Runner writes from the proposal's own fields; never text the filer wrote), waitingMinutes, expiresAt, needsSecondFactor and approvalUrl (its page in Runner). This tool never approves or rejects anything. To decide one, call approvals.request_decision with its proposalId and versionNumber, and give the person the link it returns; they confirm in Runner with their passkey. Read-only.
attention_today
ChatGPTAnswers 'what needs my attention today?' for one product: decisions waiting on an approval (count plus the ten oldest, with a link to decide each), conversations waiting on a reply or past their due time (count plus the ten most pressing), and the work queue -- overdue and unowned next actions, contacts whose consent is unknown, audience rows waiting on a policy review. Each section reports its own state, so one unreadable source never hides the rest. Links are app paths to open in Runner. Approvals are listed for the person behind this connection. Does not return message content, contact names or approval descriptions, and does not approve or change anything. Call projects.list first for the projectId.
business_overview
ChatGPTAnswers "how is the business doing?" for one product: visits, new contacts, new deals, deals closed won and revenue received over the last 7 days against the 7 days before; whether each ad channel is connected, its active campaigns, yesterday's reported spend and change plans awaiting approval; and what is waiting on a person (campaigns awaiting approval, overdue or unowned next actions, contacts with unknown consent, audience members needing a policy review). highlights are up to 5 plain sentences computed from those numbers. The window is fixed at 7 days. previous is null when no earlier week is recorded (never a guessed 0); a section that could not be read is unavailable and named in notes. Revenue is per currency, never converted or summed across currencies. Does not cover ad performance detail (use ads.overview), website traffic sources (analytics.traffic_summary) or growth targets (growth.scoreboard). Read-only.
campaigns_get_email_design
ChatGPTReads one campaign's email: its status, whether it can still be edited (a draft, in review, or approved and not yet sent), the newest message version's number and subject, and its design — a theme and an ordered list of blocks, each with an id — or null when the email is plain text or hand-written HTML (format says which). placeholderLinkBlockIds names the blocks still linking to https://example.com/. Use campaigns.list for the campaignId. Read-only, and no recipient is returned.
campaigns_list
ChatGPTLists every campaign in one project, most recently created first, scoped to the caller's organization and to a project the caller is permitted to see. Backed by lib/campaigns/campaign-queries.ts -- the same read the Campaigns page and project overview use, not MCP-only data. Returns id, name, purpose, status, kill/approval timestamps, and audience/delivery counts; no message content.
campaigns_list_email_templates
ChatGPTLists the starter designs a campaign email can begin from: each one's key, name and what it is for. Pass a key to campaigns.update_email_design as templateKey to start a campaign's email from it, in the product's own logo and colours. Read-only.
channels_list
ChatGPTLists every place product projectId can announce something, for drafting a launch or feature announcement everywhere at once: News on the product's site (Runner's News, its website feed or its own blog), LinkedIn, Instagram, Facebook, Threads, X, Bluesky, Mastodon, Reddit, TikTok, YouTube, Pinterest, Google Business Profile, Telegram, Naver Blog, Kakao Channel and the email newsletter. Each channel: connection (connected, needs_attention, not_connected, or unknown when it could not be read), growthTeamLead (the product's growth team has an active lead there, so growth.brief_posts can hand it the post), draftTools (each MCP tool that drafts for it, with channelArg, the value that tool takes for this channel; goesOut, on_approval when Runner sends it once a person approves it or hand_off when a person posts the approved words by hand, as on X, Reddit, Naver Blog and Kakao Channel; and usable, whether that tool's draft can go out now: the channel connected, or a lead on the team), connectUrl (the Runner page that connects it, when it is not connected) and note. Every draft tool files a draft or a proposal that waits for a person in Runner; none publishes. Reads only; no account name or handle is returned.
contacts_search
ChatGPTFinds contacts of product projectId whose name or email contains query (at least two characters; a full email address finds that one person). Answers each one's contactId, title, company (accountName) and whether it is archived: never a name, email or phone, which stay in Runner. Use the contactId with opportunities.add_note, or to tell whether a lead already exists before contacts.create. limit 1-25, default 10. Read-only.
content_schedule
ChatGPTAnswers 'what's scheduled to post?' and 'what went out?' for one product: posts scheduled in the next windowDays (7, 14 or 28; default 14) across every connected channel, posts due in the past windowDays with whether each published, failed, was held or missed, and recent LinkedIn company-page posts with a short excerpt. States match the content calendar. At most 100 posts per window; truncated says when there were more. Does not return engagement figures, drafts awaiting approval (attention.today lists those), or anything about audiences. Call projects.list first for the projectId.
conversations_search_unlinked
ChatGPTType-ahead search over a project's unlinked conversations (conversations not yet associated with an account/contact/opportunity), scoped to the caller's organization and to a project the caller is permitted to see. Backed by lib/conversations/unlinked-conversation-search.ts -- the same search the Contact 360 "Link existing conversation" flow uses, not MCP-only data. Returns id and a human-readable label (source + message count); no message content.
deal_rooms_ask
ChatGPTAsks the colleagues' agents in one account's deal room a question, as the person behind this connection through their own agent in that room (they must have joined it in Runner). Each agent answers only from its own records and the room's documents, under the same sharing policy as in Runner: a shared answer names its source, a held one says the owner can be asked, and the rest is refused. Nothing is generated, and nothing is posted into the room; the exchange is recorded in the sharing ledger like every exchange.
deal_rooms_read
ChatGPTReads one account's deal room in one product, as far as the sharing rules let an assistant see it: recent exchanges with each agent's answer, where an answer whose facts the rules keep from assistants has no text and a held fact appears only as 'the owner can be asked'; the plan's steps and the documents by side, status and audience; and what guest companies' teams and agents did there. Text a person typed (questions, titles, an owner's direct answer) stays in Runner. The projectId comes from projects.list and the accountId from accounts.search. Read-only: it changes nothing.
growth_play_grades
ChatGPTAnswers "which growth plays fit this product, and what was measured when other businesses ran them?" for one product: each motion from Runner's motion library that fits now (ready, or needs with how many things it waits on) or later as sales grow, with its grade: unproven, tested, not_working, moderate, strong, validated or causal. Two things are measured: the share of a line's drafts a person approved on version 1, against the same product's lines without the motion, and whether an email subject test ran. Neither measures revenue. tested says a subject test ran, not that the motion worked; not_working says most businesses did worse; causal needs a holdout and no current source awards it. organizations is a band ("5+", "10+", "25+"), never a count, null when unproven, and likeYours says whether those businesses sell the way this product does. gradedAt is when the grade was computed. A grade says what happened, never a promise. Does not cover the team's own results (growth.scoreboard). Read-only.
growth_scoreboard
ChatGPTAnswers "how are we tracking against our growth target?" for one product: its growth team's status (state is no_growth_team when the product has none); the target (what it counts, the goal, the date) with the latest actual reported toward it, how many remain and the percent reached; the last 8 weekly reviews, newest first, each with arrivals (sessions tagged from the team's channels), sign-ups, paid subscriptions, reads and revenue by currency; and what each line (channel) brought in across those weeks. Actuals only, as recorded: nothing is forecast, a week that could not be measured is unavailable with null figures (never 0), and revenue is per currency in major units, never converted. Does not cover traffic (analytics.traffic_summary), ads (ads.overview) or the overall weekly snapshot (business.overview). Read-only.
growth_team_read_seat
ChatGPTReads one seat (seatKey, from growth_team.roster) of product projectId's growth team: its name, kind, channel, status and brief (a channel lead's standing instructions; null for any other seat), and every rule that applies to it, newest first. Each rule has ruleId (what growth_team.change_rule and growth_team.retire_rule take), body, forWholeTeam (true when it is for every seat, false when it is this seat's own), source (page: the memory page; seat_thread: said in Team chat; revision: an edit on a draft; connector: added through Claude; chief_executive: added by the team's CEO agent), reason (why it was added or last changed, when one was given), addedByYou, addedAt, editedAt and readByDrafts (false when it is past the newest draftReadLimit the seat's drafts read). A person's own notes are never listed. Read-only.
growth_team_roster
ChatGPTLists every seat on product projectId's current growth team: seatKey (what growth_team.read_seat, growth_team.add_rule and growth_team.rewrite_brief take), the runner's name, kind (chief_executive, growth, finance, trust, copy, scout, outreach, channel_lead or advisor), channel (a channel lead's, else null), status (active, paused, or not_launched before the team starts) and brief, a channel lead's standing instructions (null for every other seat, which goes by its rules). ruleCount is how many rules apply to the seat, its own and the whole team's; rulesReadByDrafts is how many of those its drafts read, because a draft reads only the newest draftReadLimit. wholeTeamRuleCount counts the rules for every seat. memoryUrl is the team memory page in Runner. Read-only. Call projects.list first for the projectId.
incidents_list
ChatGPTLists the incidents in product projectId's register, newest first: incidentId, reference (INC-<year>-<n>), title, severity (sev1 worst to sev4), status (investigating, identified, monitoring or resolved), category, startedAt, resolvedAt, whether it is published to customers, and incidentUrl, where it opens in Runner. show: open (the default, everything not resolved), resolved, or all. limit: 1 to 25, default 10. Read-only; record one with incidents.record and change one with incidents.update.
journeys_get
ChatGPTReads one journey's current draft, step by step: each step's type (entry_criteria, wait, human_task, branch, action, exit or goal), its label, the words of an email step, a wait in days and which steps it leads to, with whether the draft passes Runner's checks and a link to open it. An entry rule's or a branch's conditions are not returned.
journeys_list
ChatGPTLists one product's lifecycle journeys, most recently changed first: id, name, what it is for, status (draft, review, approved, active, paused or retired) and when it last changed. Use journeys.get for a journey's steps.
leads_hot
ChatGPTAnswers 'which companies are hot right now?' for one product, from records already stored in the user's Runner workspace: up to nine companies ranked by Runner's account heat band (observed, aware, warm, engaged, qualified_for_review), with the strongest signals behind each (fit, intent, timing, relationship, reachability, risk), when the company was last active, and whether a next step is planned. Also the last 30 days of lead intake for this product: inquiries accepted, refused, scored and high-value. Companies only -- no people, contact details or browsing history. Reads stored data only; it fetches nothing live and changes nothing. Heat ranks for review; it does not mean anyone may be contacted. Call projects.list first for the projectId.
media_list_photos
ChatGPTLists the photos in a product's library, newest first, as Growth DNA, Your photos shows them. Each photo comes with a link to look at it (short-lived, so open it soon), its name, alt text, caption and tags, and what it is marked for: usedFor holds "posts" or "carousels" when the growth team may put it in one. The answer also says what Pictures in your posts is set to, how many products in the catalogue have a picture that could be brought in, and the catalogue's product names, so a photo can be named exactly as the product is written. Runner never looks at what is in a photo. For each post the team writes a short description of the picture it wants, and Runner uses the photo whose name, alt text, caption and tags share the most words with it. A photo with no name can only be found by its tags, and a description that asks for the product takes the newest photo tagged product shot, whatever it shows. Use only "unnamed" to find the photos that still need a name. Reads only; changes nothing.
media_shots
ChatGPTSplits a public video file at its cuts and returns every shot with its start, end and length in seconds, plus one keyframe per shot (taken from its middle) on labelled contact sheets (20 a sheet, 4 across, each cell marked "#n time"). Use it to write an ad's shot list: read the sheets for shot size, angle, what is in frame and where the product first appears; the cut times are measured, not guessed. url must be a DIRECT https video address up to 25 MB (a YouTube, Instagram, TikTok or X page answers needs_direct_media_url). maxShots caps the answer (default and ceiling 120); truncated says a longer film was cut short. Two flat frames of similar brightness can read as one shot, and a slow dissolve may not register as a cut. Answers status ok with shots (each with its sheet and cell), sheets (signed links that expire) and durationSeconds, or status refused with reason (invalid_url, needs_direct_media_url, media_too_large, media_unreadable, no_storage, video_service_unavailable) and message.
media_transcribe
ChatGPTTurns a public video or audio file into a transcript with timed segments, using the product's own OpenAI key from Runner's vault (whisper-1). url must be a DIRECT https media address (an .mp4, .mp3, .m4a, .wav or similar), up to 25 MB; a YouTube, Instagram, TikTok or X page cannot be read here and answers needs_direct_media_url, so fetch the media file's own address first. language is an optional ISO 639-1 hint. Answers status ok with text, language, seconds and segments (start and end in seconds), or status refused with reason (invalid_url, needs_direct_media_url, media_too_large, media_unreadable, no_key, blocked, key_rejected, rate_limited, over_budget, failed) and message saying what to do. The key never leaves Runner; each run is metered to the product's AI usage.
meetings_list_for_contact
ChatGPTLists a contact's meetings within one project, most recently scheduled first, scoped to the caller's organization and to a project the caller is permitted to see. Backed by lib/meetings/meeting-queries.ts -- the same read the Contact 360 "Meetings" section uses, not MCP-only data. Returns id, state, meeting mode, and scheduled/created timestamps; no attendee, notes, or dossier content.
newsletters_get_issue
ChatGPTReads one newsletter issue: its status, each language's subject and preview, its words as plain text, and a link to open it. Use newsletters.list to find an issue's id. No recipient is returned.
newsletters_list
ChatGPTLists one product's newsletters: key, name, whether it is set up, its languages and its latest issues with their subject and status, drafts included. Pass a key to newsletters.save_draft, or an issue id to newsletters.get_issue. No subscriber is returned.
opportunities_get
ChatGPTReads one deal (opportunityId, from opportunities.search or pipeline.summary) of product projectId: name, stage (open is true for prospecting, qualified, proposal and negotiation), value with currency, expectedCloseDate, nextStep and nextStepDueOn (YYYY-MM-DD), whether you own it, the company (accountName), the people on it (contactId, title and role, primary first; never a name, email or phone) and its notes newest first (noteId, aboutContactId, createdAt, writtenByYou). A note's text is not returned; dealUrl opens the deal in Runner, where it is. To change the deal call opportunities.update; to add a note, opportunities.add_note. Read-only.
opportunities_search
ChatGPTType-ahead search for opportunities by name, scoped to the caller's organization and to a project the caller is permitted to see. Backed by lib/opportunities/opportunity-search.ts -- the same search the Contact 360 "Link existing deal" flow uses, not MCP-only data. Returns id, name, and the linked account's name (or null); no other opportunity fields.
pipeline_summary
ChatGPTAnswers 'what deals are open?' and 'what did we win or lose lately?' for one product: open deals by stage with their value per currency, open totals per currency, deals won and lost in the last windowDays (7, 28 or 90; default 28) with their value, and the ten largest open deals with stage, expected close date and whether the next step is overdue. Values are recorded amounts in major units, grouped by currency and never converted or summed across currencies. Wins and losses count stage changes recorded in the window; a deal entered already closed is not counted. Does not forecast or weight deals, and does not return deal owners, contacts or next-step text. Call projects.list first for the projectId.
pouch_get_import
ChatGPTReads an import into Pouch that this person started: status (waiting for the file, importing, done, cancelled or expired), the counts (found, chosen, already in Pouch, left out, 2FA codes left behind, imported, failed), the social networks among them, and the logins added, each as a site and a sign-in name. Never a password, a 2FA code or a note. importId is from pouch.start_import; leave it out to read this person's latest import on the product. While it is still open, importUrl is the link to finish it.
profiles_read
ChatGPTReads product projectId's profile line (line, at most 80 characters; longer, used where a channel has room; link; suggestedBy, the seat whose card set it) and one entry per channel: channel, connected (as of Runner's last read), handle, saysNow (the bio as Runner last read it; null when it has none or could not be read), readAt, state (matches, says_something_else, too_long, not_read, not_connected, waiting_for_you for a hand-off a person has not done yet, updated in the last day, no_line before a line is set), version (the words Runner would put there, fitted to the channel, or a person's own), count and limit (characters), changesBy (runner: Runner changes it through the channel's own API, today Bluesky and Mastodon; person: a person pastes it, and reason says why: no_official_api, scope_missing, not_confirmed, not_connected, reconnect). summary counts the channels and toUpdate names the ones an update would act on. profilesUrl opens Profiles in Runner. To change them, call profiles.propose_update. Read-only.
project_get_readiness
ChatGPTReads one project's current setup-progression state (created / connected / discovery_ready / configuration_proposed / commercially_ready), scoped to the caller's organization and to a project the caller is permitted to see. Derived on every read from recorded signals (repository connections and snapshots, product type, configuration proposals, stage gate, first-proof readiness, lifecycle status) by lib/project-activation -- the same ladder the product's own setup page shows. Never stored and never attested, so there is no 'set by' user and no effective-from instant. Reports how many of the five states are reached, the requirements still standing between the project and the next one, and whether any signal could not be read (in which case the count is a floor, never an over-claim).
projects_list
ChatGPTCall this first. Lists every product (project) this connection may ask about, with the projectId every other Runner business tool needs, plus its name, slug and stage. Works whether the connection covers one product or several. It does not return any figures about the products -- use business.overview for how a product is doing, analytics.traffic_summary for its traffic and synced Google Analytics, attention.today for what needs doing, or ads.overview, pipeline.summary, revenue.summary, content.schedule and leads.hot.
revenue_summary
ChatGPTAnswers 'how much revenue did we make?' for one product: gross collected, refunds, disputes, fees and net for the last windowDays (7, 28 or 90; default 28) and for the equal window before it, plus current MRR and the last three months of MRR movement (new, expansion, contraction, churn, reactivation). Every figure is in major units with its currency; currencies are listed separately and never converted or added together. From Runner's revenue ledger of recorded Stripe events -- state is not_connected when the product has recorded none. liveMode says whether the figures are live money, Stripe test mode, or a mix. Does not return customers, invoices, subscriptions or forecasts. Call projects.list first for the projectId.
talent_availability
ChatGPTLists each talent of product projectId (a character or creator booked through its door at runr.life/<slug>) with the next open production days, soonest first: date (YYYY-MM-DD in the talent's timeZone), rush (inside the card's leadDays, priced at rushPercent more) and fromAmount, the smallest package's list price that day. packages are the card's list prices (pricing from means a starting price that always asks the talent's approver) with rushAmount for a rush day. Only what the public door shows: a talent whose card is not signed (door card_not_signed) or a creator who has not confirmed they are an adult (adult_not_confirmed) shows no prices and no days, and takes no booking. talentId narrows it to one; days is how many open days to list (default 10, at most 30), moreOpenDays how many more there are. Amounts are major units with currency. To hold one for a deal room's brand, call talent.hold_day. Read-only.
talent_bookings
ChatGPTLists the bookings of product projectId's talents from the start of last month on, by day: state (held, expired, signed, booked, released or cancelled; a hold reads expired the moment its hours run out), holdsTheDate (whether the day is closed to anyone else), waitingOn (brand_signature with holdUntil, deposit, production, or nothing), heldBy (door: a brand's verified offer; team: held with talent.hold_day), company and accountId (the deal room's account), and for an ended one endedBy and endedReason. talentId narrows it to one. No booking carries the brand's address, the amount offered or a term the brand asked for. To let an unsigned hold go, call talent.release_hold with its bookingId. Read-only.
visibility_ai_mentions
ChatGPTAnswers "when buyers ask ChatGPT, Claude, Gemini or Grok, do they name this product, and who do they name instead?" for one product. Every figure is the AI mention check Runner already stored — run every weekend on a paid plan, otherwise when someone starts it on the product's AI visibility page — never a live answer: this tool asks no assistant and costs nothing. latest is the newest finished check and previous the one before it (null when there is none); scoreChange is latest.score minus previous.score. Each check has readAt (when the assistants were asked), the 0-100 score (first place weighs most; an assistant that did not answer is left out, not counted as zero), each assistant asked with its own score and currentModel (the model Runner asks for it today; a check does not store which model answered), namedRate (answers that named the product, of answers that came back) and citedRate (answers citing a page on the product's own site, of answers that returned any sources; null means not measured, never zero). questions holds each buyer question with every assistant's answer: whether it named the product and at what position, the list it gave in its own order (ranked, the product marked ours; ranks skip where a stored list kept no name), the stored cause when it left the product out (missing_page, not_indexed, third_party_sources, weak_page or unknown), and citedSources, the pages the assistant returned as sources (null when it returned none: not measured). A host beside a ranked name is an address written in the answer's text, which is not proof the assistant retrieved that page; only citedSources are. competitors ranks the other names by how many answers gave them and how many put them ahead of the product; causes counts the lost answers' causes, and fixes are the three the product's page shows. state says why nothing is listed: never_run, no_finished_run (newerAttempt says whether a check is still running, failed or was refused by the AI budget) or unavailable (the stored checks could not be read). weeklyCheck is on when the plan repeats the check every weekend. Lists are capped. Read-only.