Account_Info
ChatGPTReturn basic account context (name, plan, status) for the current session.
Actions_Execute
ChatGPTRun a marketing platform action for a client after preview; user confirms in the app. Always use Actions_List first to obtain action_id and inputs.
Actions_List
ChatGPTCatalog of marketing platform actions (inputs, risk, descriptions). Call before Actions_Execute.
Analytics_Query
ChatGPTDefault tool for analytics questions: numbers, trends, comparisons, breakdowns, and reports from connected sources (e.g. Google Ads, GA4, Google Business Profile). Pass structured arguments: clientId, startDate and endDate (YYYY-MM-DD), measures, sources, optional dimensions, filters, timeDimension, granularity, limit/offset, and order as an object like {"googleAds.date":"asc"}. Use Datasource_List for valid source IDs; use Analytics_Schema when metric or dimension names are unknown.
Analytics_Schema
ChatGPTDiscover available analytics fields across connected data sources for a client. Use this when the right metrics or dimensions are unknown, ambiguous, or need confirmation before calling the Analytics_Query tool.
Budget_Tracker
ChatGPTJepto Budget Tracker: manage advertising and marketing spend budgets in Jepto—set monetary caps, periods, and scope (all campaigns or specific campaign IDs) on connected platforms such as Google Ads. This is spend tracking against connected data sources, not general business or household budgeting. Actions: list and info for discovery and summaries; read for the full budget plus report payload when editing; create, update, and delete. Create requires clientId, datasourceId, service, title, startDate, authEmail, connectionIds, target amount, and related options (graph type, notifications, etc.). IMPORTANT: When allCampaigns=false and campaignIds is provided,ALWAYS first call Analytics_Query with:- measures: any valid metric (e.g., 'googleAds.clicks',- sources: ['googleAds'],- dimensions: ['googleAds.campaign_name'] or ['googleAds.campaign_id'],- startDate: recent date rangeMutations use preview first with an empty confirmObject. This applies to actions create, update, and delete; The user will confirm via clicking the 'Confirm' button in the MCP app.Integrators: resolve datasource and connection fields from the client's linked data sources (Datasource_List) and confirm them before creating budgets. Use Analytics_Query to justify target and which campaigns merit a budget.Decide on default values: title from platform + scope + period; target from analytics spend context; graphType from user preference or analytics trends; period from existing budgets or analytics context. Do not ask the user for routine fields.
Client_Info
ChatGPTGet structured profile and business details for a client.
Client_Knowledgebase
ChatGPTSearch stored notes and knowledge for a client by keyword or topic.
Client_List
ChatGPTList clients in the account and their ids for use with other tools.
Datasource_List
ChatGPTList connected data sources and integrations for a client (or supported integration types when clientId is omitted). Use before Budget_Tracker, KPI flows, or analytics when you need dataSourceId, connectionIds, or authEmail.
Help_Article
ChatGPTOfficial Jepto help documentation (https://www.jepto.com/help). Two modes — pass exactly one of query or url per call, never both. (1) Search: set query to a natural-language question; optional maxResults (1–10, default 5). Returns article titles, categories, short content excerpts, and canonical URLs. (2) Fetch: set url to a help URL from a prior search result to load the full article (title, categories, headings, content). Workflow for how-to answers: call with query first, pick the best matching url from results, call again with that url, then answer from the fetched content and cite the url. Do not answer detailed setup steps from search excerpts alone. If fetch returns not found, search again with different keywords.
KPI_Forecasting
ChatGPTJepto KPI Forecasting: create, list, read, update, and delete performance KPIs in Jepto—targets and thresholds on metrics from your connected analytics and advertising sources (e.g. conversions, revenue, sessions), with periods, comparison logic, and optional campaign scope. Actions: list and info for discovery and summaries; read for the full KPI record when editing; create, update, and delete. For create, supply clientId, datasourceId, service, title, startDate, authEmail, connectionIds, metricName, targetValue, targetType, targetPeriod, and targetComparisonPeriod. Metric name must match the platform: use the Analytics_Schema tool for the same client and datasource connection (each item's name field). Google Search Console KPIs may require country and dimension. Integrators: use Datasource_List and Analytics_Query when you need performance context to choose metrics and targets.
Skills_Get
ChatGPTReturns one catalog skill by skillId: summary fields plus content (SKILL.md body). Response includes skillId, name, description, agentScope, modes, version, content.
Skills_List
ChatGPTLists skills in the Jepto catalog. Returns JSON with skills[] (skillId, name, description, agentScope, modes, version) and count. Metadata only; full markdown is returned by Skills_Get.
analytics_custom_field_create
ChatGPTCreates a reusable calculated metric or dimension for a client using Jepto's existing Formula field API. Call analytics_schema first and use exact field names in the formula. The returned field can then be used by analytics_query and report_card_create.
image_upload
ChatGPTUpload a jpg or png (base64 or data: URI) to the account's public asset store and return a hosted https URL for image cards or HTML img src. Max 2 MB. Do not pass data: URIs to report_card_create — use this tool first.
memory_list
ChatGPTRead-only list of saved Jepto memories (the Memories tab): the user's own user memories (personal preferences that apply across clients) and the account's client memories (business rules, targets, definitions and decisions for one client, shared by everyone on the account). Optional clientId (16-char hex from client_list), type (user or client), and query for semantic search ranked by relevance; without query, newest first. Returns memoryRecordId, content, type, source (data_chat or mcp), dateCreated, clientId and clientName, plus totalItems for paging with limit/offset. Check here before memory_create to avoid restating a saved memory.
report_analysis_create
ChatGPTCreate one analysis item (headline + markdown you already wrote) for a view and date range. Does not call a model. Always creates a new analysis; to revise an existing analysis use report_analysis_update. Pass groundingNumbers from analytics_query totals to verify figures in content.
report_analysis_list
ChatGPTList analysis items already saved for the session view and date range (read-only; use after report_analysis_run or report_analysis_save).
report_analysis_run
ChatGPTGenerate written analysis for a report view: provisions the analysis card when needed, invokes the report-analysis agent, and by default waits for completion and returns saved analysis rows (set waitForCompletion false for fire-and-forget).
report_analysis_save
ChatGPTSave one analysis item (headline + markdown) for a view and date range. Use after analytics research or when revising narrative text.
report_analysis_update
ChatGPTReplace the headline and/or markdown of an existing analysis item by analysisId. Pass revisions from report_analysis_list (external MCP). Stale revisions → error type CONFLICT with currentRevisions; re-read report_analysis_list and retry.
report_card_create
ChatGPTCreate a static report card (markdown, html, image, analysis, or filter). markdown/html = static copy only (not for KPI totals); analysis = executive narrative; chart/table = use report_card_generate instead. For cardType filter, set controlType (e.g. date_range). Load skills_get skillId dataChatReport for the full workflow.
report_card_delete
ChatGPTSoft-delete a card by cardId (layout updates on the server). clientId is optional when the target id is provided; delete uses account scope and that id only.
report_card_generate
ChatGPTGenerate a chart or table card. Fastest path: cardConfig with cardType chart (or table), title, and data.metrics using exact names from analytics_schema — dataType and aggregation are filled from the schema automatically. Example: {"cardType":"chart","chartType":"line","title":"Impressions","data":{"metrics":[{"name":"googleBusinessProfile.impressions"}]}}. KPI: cardType chart with chartType number. Valid cardType: chart, table, markdown, html, image, filter, analysis. Complete configs skip the model (~1s). A text prompt alone also works for vague requests, but it is slower and uses model tokens (the response shows generationMode and a hint). Run analytics_query first so the field names and date range are known. Cards are persisted unplaced: external MCP clients get needsPlacement and must place via report_page_get + report_page_update (internal report-builder may auto-place). needsAttention when the data check finds no rows (fix with report_card_update, then place). Optional dateRange overrides the preflight analytics check (default: card → page → report range → last30days). disambiguateTitle defaults false — generic titles stay as requested; suggestedTitle may appear in qualityWarnings. preflightAcknowledged is kept for compatibility and changes nothing. Load skills_get skillId dataChatReport first. On failure, error includes type, message, retryable, and optional retryAfterSeconds (retry once for retryable errors).
report_card_get
ChatGPTLoad a card configuration for editing or placement.
report_card_update
ChatGPTValidate and persist card updates; requires clientId and cardId. Pass revisions from report_card_get. Pass configDelta for data/title/pageId; optional visualizationPrompt applies chart styling. To move to another tab, remove the card from the old page layout before setting pageId.
report_card_version_get
ChatGPTRead one card archive snapshot (stored data JSON). clientId, cardId, version (>= 1). Valid versions: 1 <= version < report_card_get.revisions. Not report Change History.
report_card_version_restore
ChatGPTCard undo: restore live card from archive version (chart editor). Requires report_card_get revisions and version < revisions (one step: version = revisions - 1). Not report_version_restore. Re-read report_card_get on CONFLICT.
report_create
ChatGPTCreate empty report page shells (not a finished report). Jepto assigns viewId and pageIds. Pass every tab in pages: they are all created in this call, so do not call report_page_create for them (that duplicates tabs); use report_page_create only to add a tab to an existing report later. After report_create, run datasource_list and analytics_query (rowCount > 0) before chart report_card_generate. Then report_card_generate (repeat per card) and report_page_update per page. pages: JSON array of {title, description, layout?, dateRange?, includeDateFilter?} or JSON array string; omit for one default page. MCP adds a full-width HTML page title header on every tab. Date-range filter cards are added only on data-driven tabs (chart/table/analysis) unless includeDateFilter is true; set includeDateFilter false on static-only tabs (markdown/html/image). Response includes dateRangeRequested, dateRangePersisted, and dateRangeFilterCards. Per-page dateRange in pages[] overrides report dateRange when set. Name the first tab Executive Overview (not Page 1) for multi-page dashboards. Tabs for sources not in datasource_list are omitted (see pagesOmitted). Response includes persisted pages, pageCount, editorUrl (in-app edit), previewUrl (in-app preview), and publicUrl ({origin}/{reportId}/page/{pageId} on verified brand domain or dev|stg|jep.to) when configured. dateRange: ISO pair or preset (e.g. last30days) or JSON array string. Load skills_get with skillId dataChatReport before building. Card types: markdown/html = static notes; analysis = narrative tied to data; chart/table = metrics from analytics_schema (use report_card_generate). Response includes nextSteps and skillIds on success.
report_delete
ChatGPTSoft-delete a report by viewId. clientId is optional when the target id is provided; delete uses account scope and that id only.
report_get
ChatGPTLoad report structure, pages, layout slots, and cards for a view. Requires clientId and viewId. May include editorUrl, previewUrl, and publicUrl when link origins are configured. Layout entries list cardId and grid only — use report_card_get for card details. Use report_page_get for a single tab. On success, checklist flags empty tabs, unplaced cards, tabs without data cards, and executive/summary tabs missing an analysis card.
report_list
ChatGPTList reports (views). Omit clientId for all reports in the account; pass clientId to scope to one client. Returns viewId, clientId, title, description, dates, page summaries, and optional editorUrl. Use limit and offset for pagination; totalItems when the API returns it. Call report_get with viewId and that row's clientId for full layout and cards.
report_page_create
ChatGPTCreate a page on an existing report view. Provisions a full-width HTML page title header. Adds a date-range filter card when the page will have chart/table/analysis cards (or when includeDateFilter is true). Set includeDateFilter false for static-only tabs. Report dateRange comes from report_create / report_update. Use analysis cards on Executive/Summary/Overview tabs (report_analysis_run), not markdown with hand-typed metrics.
report_page_delete
ChatGPTSoft-delete a report page by pageId. clientId is optional when the target id is provided; delete uses account scope and that id only.
report_page_get
ChatGPTLoad a report page (title, description, layout, revisions).
report_page_update
ChatGPTUpdate page title, description, and/or layout. Use report_update for report-level dateRange (filter cards read view settings). Requires clientId and revisions from report_page_get. Send the complete layout array (every card slot on the page, not only new cards). Grid: 24 columns; x,y,w,h are grid units (each row unit h is 10px in the editor). Typical sizes (w×h): number KPI 6×32, line/bar/area 12×42, column/scatter 12×38, table 24×53, pie/donut 8×43. Slots must not overlap. To move a card to another tab: remove it from the source page layout, report_card_update with configDelta.pageId, then report_page_update on the target page.
report_palette_create
ChatGPTCreate an account palette (label plus light and dark hex color arrays).
report_palette_delete
ChatGPTDelete a palette by paletteId. Use report_palette_get to confirm what you are deleting.
report_palette_get
ChatGPTLoad one palette by paletteId (label, light and dark colours). Palette updates and deletes do not need revisions.
report_palette_list
ChatGPTList account color palettes (for themes and per-card paletteId).
report_palette_update
ChatGPTUpdate a palette patch (label, light, dark). Use report_palette_get to check current values first.
report_theme_create
ChatGPTCreate a report theme (label, optional paletteId, optional themeInput). themeInput keys: colors, grid, typography, card, maxWidth, padding, paletteId, label. See the branding workflow resource for a full themeInput example; report_theme_get returns resolved settings.
report_theme_delete
ChatGPTDelete a theme by themeId. Use report_theme_get to confirm what you are deleting.
report_theme_get
ChatGPTLoad one theme by themeId. Theme updates and deletes do not need revisions.
report_theme_list
ChatGPTList account report themes (optional viewId filter).
report_theme_update
ChatGPTUpdate a theme (patch and/or themeInput). Use report_theme_get to check current values first.
report_update
ChatGPTUpdate report metadata (title, description, date range, theme).
report_version_get
ChatGPTHydrated Change History snapshot (config only). clientId, viewId, version (0=current, 1+=historical). pages/cards are top-level. Manifest revisions from version 0 for report_version_restore — not report_get.revisions.
report_version_list
ChatGPTChange History timeline for a view (manifest versions). Requires clientId and viewId. Returns changeGroups and pagination; currentRevisions is manifest lock from isCurrent on this page only — use offset 0 before restore. Card undo: report_card_version_restore.
report_version_restore
ChatGPTRestore full report from Change History (version >= 1). Requires manifest revisions from report_version_list (offset 0) or report_version_get(version: 0) — not report_get. Re-list on CONFLICT; then report_get. For single-card undo use report_card_version_restore.
view_card
ChatGPTShows one Jepto chart or table card inline as an interactive MCP App, rendered as in Jepto with live data. Takes clientId and the cardId from report_card_create or report_get. Works for exploration cards (created with viewId view_0000000000000000 and pageId page_0000000000000000, not on any report) and for cards on a report; only the card is loaded, never the report. After report_card_update, call view_card again to show the new version. Uses a one-hour viewer token scoped to this card.
view_report
ChatGPTShows a Jepto report inline as an interactive MCP App: the same page tabs and live cards as the report's share link. Takes the viewId from create_report or get_report and an optional pageId to open first. Returns the share URL, share id, title and access. Reports whose share link is private or disabled return the URL only.
web_fetch
ChatGPTFetch and extract content from a public URL (not report-specific).
web_search
ChatGPTSearch the public web for external context (not report-specific).
Account_Info
ClaudeActions_Execute
ClaudeActions_List
ClaudeAnalytics_Query
ClaudeAnalytics_Schema
ClaudeBudget_Tracker
ClaudeClient_Info
ClaudeClient_Knowledgebase
ClaudeClient_List
ClaudeDatasource_List
ClaudeHelp_Article
ClaudeKPI_Forecasting
ClaudeSkills_Get
ClaudeSkills_List
Claude