{"name":"Versant Mega Tracker Agent","description":"Answers questions over the Versant Mega Tracker brand-health survey for CNBC, MS NOW, USA Network, Golf Channel, Fandango, Fandango at Home, Rotten Tomatoes, E!, Oxygen, Syfy, and the Versant golf-app cluster. Returns funnel KPIs, brand affinity, NPS, perception attributes, and multi-wave trends with audit-ready citations.","supportedInterfaces":[{"url":"https://versant-agent.thefocus.ai/api/a2a/jsonrpc","protocolBinding":"JSONRPC","protocolVersion":"1.0"}],"provider":{"url":"https://thefocus.ai","organization":"The Focus AI"},"version":"0.1.0","documentationUrl":"https://versant-agent.thefocus.ai/docs","capabilities":{"streaming":true,"pushNotifications":false,"extensions":[{"uri":"urn:versant:refusals/v1","description":"KPIs the agent will not compute. Callers should not request these; the agent will refuse with a one-line explanation if asked anyway. Reasons mirror app/agent/refusals.ts and the system prompt — single source of truth.","params":{"seg-mix":"The four segmentation variables (PolarisSegment, POLARIS_SUBSEGMENT_2, MSNBC Segment, CNBC_TOOL_SEGMENT_ASSIGNMENT) have similar names but different value spaces — combining them produces nonsense bases. Pick one segmentation column explicitly.","invalid-cut":"A cut's value isn't one of the supported options for its dimension. Cut values are exact strings — see the catalog enumerated in the tool description, or call listAvailableCuts() if exposed."}},{"uri":"urn:versant:data-source/v1","description":"Identity of the underlying survey and its sampling frame.","params":{"study":"Versant Mega Tracker","vendor":"National Research Group","universe":"U.S. Adults 18-74","sampleSize":"n≈8,000 per wave, ~24,000 per rolling-quarter","fielding":"Monthly since November 2025"}}]},"securitySchemes":{"clerkOAuth2":{"oauth2SecurityScheme":{"description":"Client-credentials grant. Register a Clerk M2M application and exchange client_id + client_secret for a short-lived JWT.","flows":{"clientCredentials":{"tokenUrl":"https://model-rabbit-82.clerk.accounts.dev/oauth/token","scopes":{"versant:read":"Read access to brand KPIs, perceptions, trends, and competitive comparisons."}}}}}},"securityRequirements":[{"schemes":{"clerkOAuth2":{"list":["versant:read"]}}}],"defaultInputModes":["text/plain"],"defaultOutputModes":["text/plain","application/json","text/html+mcp","application/vnd.versant.citations+json"],"skills":[{"id":"getBrandFunnel","name":"Brand Funnel","description":"Aided awareness, familiarity, and brand-user share for one focal brand × one wave. Source: S14. Weight: main_weight.","tags":["funnel","awareness","familiarity","users"],"examples":["What's the aided awareness of CNBC in Wave 4?","How familiar are people with Rotten Tomatoes?"]},{"id":"getBrandPerceptions","name":"Brand Perceptions and NPS","description":"Top-2-Box perception attributes plus brand affinity and Net Promoter Score for one focal brand × one wave. Track-aware base + scale handling.","tags":["perception","nps","affinity","t2b"],"examples":["What are MS NOW's brand perceptions in Q1 26?","What's CNBC's NPS among current users?"]},{"id":"getTrend","name":"Multi-Wave Trend","description":"Trajectory of any KPI across a range of waves for one brand. Supports rolling-quarter aggregation.","tags":["trend","time-series","rolling-quarter"],"examples":["How has Fandango familiarity trended W3-W5?","Show me MS NOW NPS over the last three months."]},{"id":"chartSeries","name":"Generic Series Chart","description":"Line chart over waves for series assembled from prior tool results — charts anything outside the S14 metric set (loop rows, distributions, multi-select options). Presentation-only; values come from the underlying data tools' citations. Ships CSV and PNG export.","tags":["chart","trend","time-series","export"],"examples":["Line graph of all MVPD providers' likelihood to leave across the five waves."]},{"id":"calculate","name":"Calculate","description":"Sum, difference, ratio, percent-of, mean, or weighted mean over cells other tools returned this turn. Every input names its source tool; the tile prints the inputs, the written arithmetic, and the result rounded once at the end. The agent does no arithmetic in its own text.","tags":["arithmetic","derived","t2b","ratio","average"],"examples":["What share of people who ever used Fandango have lapsed?","Average CNBC and MS NOW trust as an ad-hoc set."]},{"id":"getOneSheet","name":"Published One-Sheet","description":"Reproduces a published Versant one-sheet PDF: funnel + perceptions block for one focal × one wave. Convenience wrapper.","tags":["one-sheet","report"],"examples":["Give me the W4 CNBC one-sheet.","Show me Fandango at Home for Wave 3."]},{"id":"getCompetitiveFunnel","name":"Competitive Funnel","description":"Funnel KPIs for the focal brand alongside every competitor in its published competitive set. Returns one row per brand.","tags":["competitive","comparison","funnel"],"examples":["How does Rotten Tomatoes compare to IMDb and Letterboxd on familiarity?"]},{"id":"getCompetitiveMetric","name":"Competitive Metric","description":"A single KPI (NPS, T2B affinity, a specific perception attribute) across the focal + competitive set. Powers the deck's competitor bar charts.","tags":["competitive","comparison","nps","t2b"],"examples":["Compare CNBC's 'Is a brand I trust' to its comp set."]},{"id":"getOpinionBands","name":"Opinion Bands","description":"Three-way Love/Like / Somewhat / Hate-Dislike breakdown of brand affinity. Reproduces the stacked-bar slides in quarterly decks.","tags":["affinity","opinion","bands"],"examples":["Show me the love/like/hate breakdown for MS NOW in Q1 26."]},{"id":"getBrandSatisfaction","name":"Overall Satisfaction","description":"All five overall-satisfaction levels (E10B / N12B / SP14B) for one brand, with the base n and the top-2 and top-3 nets. Fielded from W9 (July 2026). Pooled over waves or month by month.","tags":["satisfaction","levels","top-2"],"examples":["Break down each satisfaction level for Rotten Tomatoes in W9.","How satisfied are CNBC's current users, month by month?"]},{"id":"getFrequency","name":"Frequency of Use","description":"Distribution of N11/E*/SP* frequency-of-use codes for a brand × wave.","tags":["frequency","engagement"],"examples":["How often do current CNBC users engage with the brand?"]},{"id":"getMultiSelectDistribution","name":"Multi-Select Distribution","description":"Binarized %-selected for multi-select grids: N15 brand traits, N18 platforms, N23 engagement drivers, N24 barriers, E13, SP20, SP21.","tags":["multi-select","traits","drivers","barriers"],"examples":["What attributes describe MS NOW for current users?","What platforms do CNBC users consume the brand on?"]},{"id":"applyCut","name":"Demographic Cut","description":"Reruns any of the above with a demographic filter (gender × age × ideology × segment). Composable: pass multiple cuts and they AND together. Warns automatically on small bases.","tags":["demographics","cut","filter","subgroup"],"examples":["Show me CNBC funnel for female 25-34 Liberals.","How do Polaris-segment news consumers rate MS NOW?"]},{"id":"listAvailableBrands","name":"List Available Brands","description":"Discovery: returns the focal brands the agent can answer about, grouped by track. Use when the user mentions a brand we may not have configured.","tags":["discovery","metadata"],"examples":["What brands can you tell me about?"]},{"id":"listAvailableWaves","name":"List Available Waves","description":"Discovery: returns which survey waves are currently loaded. Use before answering trend questions.","tags":["discovery","metadata","waves"],"examples":["What time periods do you have data for?"]},{"id":"getMethodologyNote","name":"Methodology Note","description":"Returns the methodology disclosure for any KPI — base definition, weight column, scale, T2B rule, known caveats. Use when a user questions a number.","tags":["methodology","citation","disclosure"],"examples":["How is NPS calculated here?","Why is the affinity base different on slide 13?"]}]}