Pendo MCP

[STALE: upstream tool "accountQuery" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; may have been removed or replaced.] Retrieve account data and metadata, or count matching accounts.

[STALE: upstream tool "activityQuery" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; may have been removed or replaced.] Query aggregated activity metrics for pages, features, and track types over a date range.

Lists and ranks individual AI agent conversations with per-conversation metrics. Returns one row per conversation with: conversationId, visitorId, accountId, startTime, numRagePrompts, numErrors, and firstPromptContent. Supports filtering to a specific set of conversations and sorting by rage prompt count, error count, or date. USE FOR: Drilling down from aggregate metrics to individual conversation-level investigation - surfacing which specific conversations are driving a metric such as a high rage-prompt rate. When conversationIds are provided, the list is narrowed to a specific cluster (e.g. conversations associated with a detected issue or a use case). EXAMPLES: - Show me the conversations with the most rage prompts for my chat agent - Which conversations had the most errors this week? - List the most recent conversations for agent X - Show me these specific conversations: [id1, id2, id3] NOT FOR: Aggregate volume metrics for an agent. Diagnosing an issue cluster (explanations, tools invoked, response patterns). Discovering or clustering use cases. RETURNS: - [conversations]: one row per conversation with conversationId, visitorId, accountId, startTime, numRagePrompts, numErrors, firstPromptContent. The maximum time range for this tool is 90 days.

Returns key aggregate metrics for AI agent conversations with period-over-period comparison. Includes: conversations, visitors, accounts, prompts, rage prompt rates (per-prompt and per-conversation), visitorIds, accountIds, and visitor retention. All metrics include previous-period equivalents for trend analysis. USE FOR: Getting a high-level summary of AI agent usage and engagement. Use when the user asks about overall agent performance, conversation volume, visitor engagement, rage prompt rates, or retention for a specific agent. EXAMPLES: - How many conversations has my ABC agent had in the last 30 days? - What is the rage prompt rate for Acme agent this month? - Show me key metrics and trends for my chat agent over the past 2 weeks. - How many unique visitors have used my chat agent recently? NOT FOR: Use listUseCases for topic/cluster analysis. Use listAiAgentIssues for issue detection. Use listAiAgents to get agent IDs and names first when the user has not specified an agent. RETURNS: - Current period: numConversations, numPrompts, numVisitors, numAccounts, numRagePrompts, ragePromptsRate, ragePromptsConversationRate, retention (retentionRate, retained, visitors), visitorIds, and accountIds. - Previous period (same duration, immediately prior): prevNumConversations, prevNumPrompts, prevNumVisitors, prevNumAccounts, prevNumRagePrompts, prevRagePromptsRate, prevRagePromptsConversationRate, prevRetention. - Also includes conversationsWithRagePrompts and prevConversationsWithRagePrompts. The maximum time range for this tool is 90 days.

Deep-dives into a single tracked use case in AI Agent Analytics for a specific AI agent, surfacing the visitors and accounts, the tools and models the agent invoked, sampled explanations, and a sample of the user prompts associated with the tracked use case. USE FOR: Deep-diving into a specific tracked use case - call directly if agentId is known and conversationIds, eventIds, and trackedUseCaseId are passed. Use when the user wants to understand what prompts users sent for a tracked topic, which tools and models the agent invoked, or which visitors and accounts are associated with a tracked use case. EXAMPLES: - What are users actually asking in the 'dashboard help' tracked use case? - Which tools does my agent use when handling this tracked use case? - Who are the visitors in this tracked use case? - Show me sample prompts from this tracked use case. NOT FOR: Aggregate volume metrics across use cases, or issue diagnosis - this tool deep-dives a single tracked use case. RETURNS: - Four labeled CSV sections: - [tracked_use_case_summary]: visitorIds, accountIds - one summary row with deduplicated lists. - [explanations]: explanation - one row per sampled explanation (capped for MCP size). - [tools_used]: toolsUsed, modelsUsed - one summary row with deduplicated lists. - [prompt_samples]: content - one row per sampled user prompt message (capped for MCP size). The maximum time range for this tool is 90 days.

Rank guides by aggregate usage over a date range, returning one row per guide. Where guideMetrics analyses a single known guide in depth, this tool compares an entire cohort's usage across every guide. Each row has entityId, entityName, appId, totalViews, totalCompletions, totalDismissals, uniqueVisitors, uniqueAccounts, and viewsPerUser. totalViews excludes continue-resumed guideSeen events to match the guide-details UI. Deleted guides surface with entityName "(Deleted Guide)". USE FOR: Cross-guide ranking - e.g. top guides by views, which guides have the most dismissals, least-used guides. EXAMPLES: - Top 10 guides by views last month - Which guides have the most dismissals? - Show me the least-completed guides over the last 30 days - Rank guides by unique visitors this quarter - Top public tooltips by views - Compare views for these three guides RETURNS: - meta.dateRange: resolved startDate and endDate (YYYY-MM-DD) - rows: one per guide with entityId, entityName, appId, totalViews, totalCompletions, totalDismissals, uniqueVisitors, uniqueAccounts, viewsPerUser FILTERING OPTIONS: - guideIds: restrict the ranking to specific guide IDs - status: guide state - public, staged, scheduled, draft, pendingReview, inactive - guideType: guide type - banner, tooltip, lightbox, walkthrough, whatsnew, building-block, group, training, launcher, mobile-lightbox - activation: launch method - auto (automatic), api, badge, dom (element click), embed, launcher (resource center), page, feature, form, track - productAreaIds / guideCategoryIds: guides belonging to those product areas or guide categories - pageIds: guides whose first step is on one of those pages ("sitewide" matches guides with no page) - appId, accountId, segmentPipeline: scope the underlying guide activity - Note: any filter other than guideIds resolves guides from their current metadata, so deleted guides drop out of the results.

Get per-visitor or per-account app usage metrics for a date range. Returns {summary, rows}: summary has total active visitors/accounts, total events across the selected app scope, average daily time on apps, and totals for the four frustration counts (totalErrorClickCount, totalRageClickCount, totalUTurnCount, totalDeadClickCount); rows are per-visitor or per-account breakdowns with daysActive, totalTime and avgTimePerDay (duration objects {seconds, display}), totalEvents, and the same four frustration totals, sorted and limited. When groupBy="account" each row also includes numVisitors - the count of distinct visitors from that account who were active during the window. Optionally include metadata fields for each returned visitor or account with select. Audience can be scoped via an inline segmentPipeline; scope can also be narrowed to a single app via appId, or to a single product area via productAreaId. When productAreaId is supplied the same summary and rows describe headline usage metrics for that product area, measured over its page, feature, and track-event activity. USE FOR: Per-visitor or per-account app-level usage (events, time, days active) AND top-N ranking over a window across every app, or one app when appId is supplied, or one product area when productAreaId is supplied (headline usage metrics for that area). EXAMPLES: - Who uses our apps the most in the last 30 days? - Top 50 accounts by app usage last quarter - Which visitors rage-click the most across our apps last week? - Top 20 visitors by total time in app X last month - Headline usage metrics for product area X over the last 30 days - Top 20 accounts by usage of product area X last quarter - Show the top 20 accounts by app usage last quarter with their ARR and company size RETURNS: - meta.dateRange: resolved startDate and endDate (YYYY-MM-DD) - summary: totalNumVisitors or totalNumAccounts, totalNumEvents, avgDailyActiveTime (duration object {seconds, display}), totalErrorClickCount, totalRageClickCount, totalUTurnCount, totalDeadClickCount - rows: daysActive, totalTime, totalEvents, avgTimePerDay, totalErrorClickCount, totalRageClickCount, totalUTurnCount, totalDeadClickCount per visitor or account; per-account rows also include numVisitors; select adds requested metadata fields

buildPendoSegment: Purpose: Describe one visitor segment by visitor or account ID, activity on Pages, Features, TrackEvents, Guides, the elements inside a guide, segment membership, or metadata. Input shape: pass definition as an array of rule groups. Top-level groups are ANDed together; rules within a group are ORed together. For one ANDed rule, wrap it in a single-item group. SINGLE CALL CONTRACT (important): The ENTIRE segment, including every AND clause and every OR clause, MUST be built with ONE call to this tool. Never split a segment across multiple calls. Combine all clauses into a single definition array and pass it once. Calling the tool more than once produces multiple unrelated segments, NOT one combined segment, so the AND/OR logic between the calls is lost. Mapping a request to a single definition: - Split the request into top-level AND clauses. - Each AND clause becomes one group (one inner array) in the definition. - Inside a clause, every OR alternative becomes one rule object in that group. - A clause with no OR is a group containing a single rule object. CLAUSE COVERAGE (important): Account for EVERY clause in the request. Before returning, count the AND clauses in the request and confirm the definition has exactly that many top-level groups, then confirm every OR alternative inside each clause is present as a rule object. Never drop, merge, or skip a clause. If the request mentions a page rule, a feature rule, and a track event rule joined by AND/OR, all of them must appear. A request like "(X OR Y) AND Z" has 2 top-level groups (one for "X OR Y", one for "Z") and 3 rule objects total; returning fewer groups or rules is wrong. Full example (combining AND and OR in one call): Request: "(visitors who spent > 15 minutes on page A in May OR have NOT seen page B in the last year) AND triggered track event C on fewer than 5 days in the last 7 days" Correct single definition (one tool call): [ [ { "entityType": "page", "entityId": "A", "metric": "eventTime", "operator": ">", "threshold": 15, "condition": "between", "first": "2026-05-01", "last": "2026-05-31" }, { "entityType": "page", "entityId": "B", "metric": "notseen", "condition": "withinLast", "lookbackAmount": 12, "granularity": "months" } ], [ { "entityType": "trackEvent", "entityId": "C", "metric": "daysActive", "operator": "<", "threshold": 5, "condition": "withinLast", "lookbackAmount": 7, "granularity": "days" } ] ] The first group (the two page rules) is ORed together; that group is then ANDed with the second group (the track event rule). This is ONE call, not two. REQUIRED FIELDS AND CONDITION CHOICE (important): Activity metrics with non-empty supportedConditions MUST include an explicit "condition" field. It is the discriminator that selects which time/frequency fields apply, and it is NOT optional for those metrics. Always choose condition from the chosen metric's supportedConditions. Do NOT rely on the presence of first/last, lookbackAmount/granularity, or date to imply the condition; you must still set condition explicitly for activity metrics. For example: - first + last present => you MUST also set condition: "between" - lookbackAmount + granularity => you MUST also set condition: "withinLast" - date present => you MUST also set condition: "since" Segment membership metrics are the exception: - isMemberOfSegment and isNotMemberOfSegment have supportedConditions: [] by design. - For these metrics, omit "condition" entirely. Do NOT add "ever" or infer any default condition. - supportedConditions: [] means "no condition field", not "choose a default condition". Visitor and account ID rules are conditionless and metricless: - Use entityType: "visitor" or "account". - Use operator: one of "==", "!=", "contains", "!contains", "empty", "!empty". Omit it to default to "==". - Use entityId for the comparison value with "==", "!=", "contains", and "!contains". - Use an empty entityId with "empty" and "!empty". - Example: { "entityType": "visitor", "entityId": "bob@example.com" }. - Example: { "entityType": "visitor", "entityId": "@example.com", "operator": "contains" }. - Example: { "entityType": "account", "entityId": "acme-corp" }. Metadata rules are also conditionless and metricless: - Use entityType: "metadata". - Use entityId for the full metadata key, e.g. "visitor.agent.job_role". - Use operator: one of "==", "!=", ">=", "<=", "contains", "!contains", "empty", "!empty". - Use value for "==", "!=", ">=", "<=", "contains", and "!contains". Omit value for "empty" and "!empty". - Time and date metadata values may use ISO 8601, e.g. "2025-12-31T23:59:59Z". - Example: { "entityType": "metadata", "entityId": "visitor.agent.job_role", "operator": "==", "value": "Software engineer" }. Guide element rules cover clicks on one button, link or close icon inside a guide: - Use entityType: "guideElement". - Use entityId for the GUIDE id, stepId for the step the element sits on, and elementId for the element's uiElementId, e.g. { "entityType": "guideElement", "entityId": "<guideId>", "stepId": "<guideStepId>", "elementId": "pendo-button-a1b2c3d4", "metric": "eventCount", "operator": ">=", "threshold": 1, "condition": "withinLast", "lookbackAmount": 30, "granularity": "days" }. - One getEntity call gives both ids: steps[].stepId and the steps[].elements[].uiElementId nested under it. Never guess either. - eventCount is the only supported metric, and it counts clicks on that element in the chosen window: - "clicked it" => operator ">=", threshold 1 - "did not click it" => operator "==", threshold 0 - click counts => any operator, e.g. "clicked it more than 3 times" is operator ">", threshold 3 - stepId does not narrow the count. - Only withinLast and between are available, so a click rule is always scoped to a window. Resolution steps 1. Choose entityType from EntityTypes. 2. For visitor or account, use entityId + optional operator and stop here. 3. For metadata, use entityId + operator + optional value and stop here. 4. For activity and segment entities, choose metric from EntityTypes[entityType].supportedMetrics. 5. Look up Metrics[metric]. 6. If Metrics[metric].supportedConditions is non-empty, choose condition from that list and look up Conditions[condition]. 7. The final rule fields are: EntityTypes[entityType].requiredFields + Metrics[metric].requiredFields + Conditions[condition].requiredFields, only when a condition is required. Time phrase resolution examples: - "this year" / "so far this year" / "year to date" => condition: "since", date: "2026-01-01" - "this month" / "month to date" => condition: "since", date: "2026-06-01" - "since 2026-01-01" => condition: "since", date: "2026-01-01" - "last year" => condition: "between", first: "2025-01-01", last: "2025-12-31" - "in the last year" => condition: "withinLast", lookbackAmount: 12, granularity: "months" - "between 2026-01-01 and 2026-06-05" => condition: "between", first: "2026-01-01", last: "2026-06-05" - "in March" => condition: "between", first: "2026-03-01", last: "2026-03-31" - "last 3 weeks" => condition: "withinLast", lookbackAmount: 3, granularity: "weeks" - "last 7 days" => condition: "withinLast", lookbackAmount: 7, granularity: "days" CONDITION CHOICE FOR TIME PHRASES (important): - Use "since" (set only date) for any open-ended period running from a start date up to now: "this year", "this month", "year to date", "so far", "since <date>". Do NOT express these as "between ... today". - Use "between" (set first + last) ONLY for a closed range whose end is a specific past date or the last day of a named calendar period (e.g. "in March", "last year", "between X and Y"). - Use "withinLast" (set lookbackAmount + granularity) for rolling windows: "last N days/weeks/months". - Set ONLY the fields the chosen condition requires. Never combine date with first/last in one rule. A segment should be just as valid in a week's time as it is today, so never hardcode today's date as an endpoint. For a fixed/named calendar period (e.g. "in March") use "between" spanning the entire period even if it has not finished yet. For an open-ended current period (e.g. "this year", "this month") use "since" with the period's start date, which stays valid going forward. EntityTypes: visitor: requiredFields: entityType: "visitor" entityId: String supportedMetrics: [] account: requiredFields: entityType: "account" entityId: String supportedMetrics: [] page: requiredFields: entityType: "page" entityId: String supportedMetrics: ["eventCount", "deadClicks", "errorClicks", "rageClicks", "daysActive", "uTurns", "eventTime", "seen", "notseen", "lastSeen"] feature: requiredFields: entityType: "feature" entityId: String supportedMetrics: ["eventCount", "deadClicks", "errorClicks", "rageClicks", "daysActive", "used", "notused", "lastused"] trackEvent: requiredFields: entityType: "trackEvent" entityId: String supportedMetrics: ["eventCount", "daysActive", "used", "notused", "lastused"] guide: requiredFields: entityType: "guide" entityId: String supportedMetrics: ["seen", "lastSeen", "notSeen"] guideElement: requiredFields: entityType: "guideElement" entityId: String stepId: String // the guide step the element sits on elementId: String // the element's uiElementId, e.g. "pendo-button-a1b2c3d4" supportedMetrics: ["eventCount"] segment: requiredFields: entityType: "segment" entityId: String supportedMetrics: ["isMemberOfSegment", "isNotMemberOfSegment"] metadata: requiredFields: entityType: "metadata" entityId: String supportedMetrics: [] Metrics: eventCount: description: Number of events for the selected entity. requiredFields: metric: "eventCount" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] deadClicks: description: Number of dead clicks for the selected page or feature. requiredFields: metric: "deadClicks" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] errorClicks: description: Number of error clicks for the selected page or feature. requiredFields: metric: "errorClicks" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] rageClicks: description: Number of rage clicks for the selected page or feature. requiredFields: metric: "rageClicks" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] daysActive: description: Number of active days for the selected entity. requiredFields: metric: "daysActive" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] uTurns: description: Number of u-turns for the selected page. requiredFields: metric: "uTurns" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] eventTime: description: Time in minutes spent on the selected page. requiredFields: metric: "eventTime" operator: Enum["==", "!=", ">=", "<="] threshold: Integer supportedConditions: ["withinLast", "between"] used: description: Whether the selected feature or track event was used, optionally with a frequency threshold. requiredFields: metric: "used" supportedConditions: ["ever", "since", "withinLast", "atLeast", "atMost"] notused: description: Whether the selected feature or track event was not used. requiredFields: metric: "notused" supportedConditions: ["ever", "since", "withinLast"] lastused: description: When the selected feature or track event was last used. requiredFields: metric: "lastused" supportedConditions: ["since", "withinLast", "between"] seen: description: Whether the selected page was seen, optionally with a frequency threshold. For checking withinLast requiredFields: metric: "seen" supportedConditions: ["ever", "since", "withinLast", "atLeast", "atMost"] notseen: description: Whether the selected page was not seen. requiredFields: metric: "notseen" supportedConditions: ["ever", "since", "withinLast"] lastSeen: description: When the selected page was last seen. requiredFields: metric: "lastSeen" supportedConditions: ["since", "withinLast", "between"] isMemberOfSegment: description: Whether the visitor is a member of the selected segment. requiredFields: metric: "isMemberOfSegment" supportedConditions: [] isNotMemberOfSegment: description: Whether the visitor is not a member of the selected segment. requiredFields: metric: "isNotMemberOfSegment" supportedConditions: [] Conditions: ever: description: The metric happened at any time. requiredFields: condition: "ever" since: description: The metric happened on or after a specific date. requiredFields: condition: "since" date: DateString("yyyy-mm-dd") withinLast: description: The metric happened within a rolling time window. requiredFields: condition: "withinLast" lookbackAmount: Integer granularity: Enum["days", "weeks", "months"] between: description: The metric happened within an inclusive date range. requiredFields: condition: "between" first: DateString("yyyy-mm-dd") last: DateString("yyyy-mm-dd") atLeast: description: The metric happened at least threshold times ever requiredFields: condition: "atLeast" threshold: Integer atMost: description: The metric happened at most threshold times ever requiredFields: condition: "atMost" threshold: Integer

Get usage analytics for one known page, feature, or track event ID over a date range, including page views, feature clicks, event counts, and unique visitor ('people') and account counts. Returns {summary, rows}: summary has total events, unique visitors, unique accounts (and for pages/features, frustration counts; pages also include uTurnCount); rows are per-visitor or per-account breakdowns with events, time on entity, and days active, sorted and limited. Optionally include metadata fields for each returned visitor or account with select. Audience can be scoped via visitorIds OR an inline segmentPipeline (mutually exclusive), and/or filtered to a single accountId. USE FOR: Scalar 'how many people viewed/clicked/used it' totals and top-N visitor or account questions over a window for one known page, feature, or track event ID. EXAMPLES: - How many page views did page X get last month? - Top 20 visitors by time on page X in the last 30 days - Top 50 accounts by usage of feature Y last quarter - How many rage clicks did feature Y have last week? - Show the top 20 visitors by time on page X last month with their email and role NOT FOR: Looking up an entity ID from its name, or ranking multiple entities against each other. WORKFLOW: When the user names a page, feature, or track event but does not provide its ID, call listCountables to resolve the name to an ID, then call entityUsage with that ID. Do not substitute aggregateEntityUsage for this lookup: its limited ranking may omit the named entity. RETURNS: - summary: totalNumEvents, totalNumVisitors, totalNumAccounts (pages/features also: totalErrorClickCount, totalRageClickCount, totalDeadClickCount; pages also: totalUTurnCount) - rows: events, timeOnEntityInMinutes, daysActive per visitor or account (pages/features also: errorClickCount, rageClickCount, deadClickCount; pages also: uTurnCount); per-account rows also include numVisitors; select adds requested metadata fields

[STALE: upstream tool "get_agent_context" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; likely renamed to "getAgentContext".] Fetch a grounding document describing a Pendo product resource so an LLM can answer questions about it.

Returns the configuration for a given AI agent: its name, description (a human-authored summary of the agent's role and purpose, not its LLM system prompt), model preset, type, and the tool names and descriptions it has used in the specified date range. The agent config section comes from the stored agent record. The tools section is derived from runtime events in the date range and may be empty if the agent had no activity. USE FOR: Fetching the static configuration and runtime tool inventory of an AI agent. Use before analyzing issues or generating fix briefs to understand what the agent is configured to do and which tools it calls. EXAMPLES: - What is my agent's configured role and purpose? - Which tools does my agent have access to? - What model does agent X use? - Show me the full config for this agent before I debug it NOT FOR: The agent's actual LLM system prompt - this tool does not have access to that. Aggregated usage metrics, issue clusters, or conversation analysis - use the agent analytics tools for those. RETURNS: - Two labeled CSV sections: - [agent_config]: name, description (human-authored role/purpose summary, not the LLM system prompt), preset (model), type - one row. - [tools]: toolName, toolDescription - one row per tool seen in the date range; empty section if no activity found. The maximum time range for this tool is 90 days.

[STALE: upstream tool "guideMetrics" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; likely renamed to "aggregateGuideMetrics".] Get performance metrics for a single guide over a date range, including reach, views, and completion rates.

Get per-visitor or per-account usage breakdown for a single guide, with time-on-guide and new vs returning viewers. totalViews excludes continue-resumed guideSeen events to match the Pendo guide-details UI. For poll guides, includes per-poll response counts and response rate in summary. USE FOR: Per-visitor or per-account guide engagement - e.g. top visitors by time on guide, which accounts are completing vs dismissing, new vs returning viewer counts. With includeElements=true, also attributes each step's clicks to the individual buttons/links on it. EXAMPLES: - Top 20 visitors by time on guide X in the last 30 days - Which accounts are dismissing the onboarding guide most? - How many new vs returning viewers did the walkthrough have this month? - Show me per-visitor completion data for the tooltip guide - Is drop-off at step 3 happening via the CTA or by dismissing the guide? - Which button on the last step of the onboarding guide gets clicked most? RETURNS: - meta.dateRange: resolved startDate and endDate (YYYY-MM-DD) - summary: totalViews, completions, dismissals, uniqueVisitors, uniqueAccounts, completionRate, dismissalRate, viewsPerUser, avgTimeOnGuide, medianTimeOnGuide (duration objects {seconds, display}), newViewers, returningViewers, isPoll - summary.polls (poll guides only): per-pollId uniqueResponseCount and totalResponseCount - rows: per-visitor or per-account rows with views, completions, dismissals, daysActive, lastSeen (timestamp object {iso, display}), timeOnGuide (duration object {seconds, display}) (per-account rows also include uniqueVisitors), sorted and limited - steps (only when includeSteps or includeElements is true): per-step array in guide step order, each with stepId, name, uniqueVisitors, viewCount. Single-step guides return one entry; step views include continue-resumed views to match the Pendo step funnel. - steps[].elements (only when includeElements=true): the step's clicked buttons/links, highest clicks first, each with uiElementId, text, type ('Button', 'Close Button', 'Link', 'Task Item', 'Image', 'Swiped Left', 'Swiped Right' or 'Unknown'), actions (what the click does, e.g. 'Next Step', 'Dismiss Guide', 'External URL', 'Go to Step (2)'), clicks, uniqueVisitors, percentOfStepClicks. Compare an element's uniqueVisitors against its step's uniqueVisitors to attribute drop-off. Only clicked elements appear, and elements since removed from the step are still counted. The maximum time range for this tool is 367 days.

[STALE: upstream tool "list_ai_agent_issues" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; likely renamed to "listAiAgentIssues".] List detected issues in an AI agent's conversations with instance and conversation counts.

[STALE: upstream tool "list_ai_agents" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; likely renamed to "listAiAgents".] List all AI agents accessible in this subscription with agent IDs, names, and deployment configuration.

[STALE: upstream tool "list_spaces" is no longer present in the Pendo MCP server's tools/list as of 2026-08-19; likely renamed to "listSpaces".] List the Pendo Spaces accessible to the current user.

Lists detected (emergent) issues in AI agent conversations with instance counts and conversation counts. Returns a table of issue name (clusterName), summary, instance count, and conversation count per issue, plus a sample of conversationIds/eventIds for deep-diving via agentAnalyticsIssueAnalysis. It also returns visitorIds and accountIds who experienced the issue. USE FOR: Finding what problems or issues users encountered when using AI agents (e.g., incorrect answers, refusals, errors). Use when the user asks about common issues, problems, or failures with a specific agent. Refer to an issue by its clusterName, never its numeric clusterId. EXAMPLES: - What are the common issues my users are encountering using my ABC agent in the last 30 days? - What problems have users been encountering when using Acme agent in my Acme application lately? - Have we seen reduced occurrences of issues with incorrect answers in our fooBar agent since last month? - What models are most often used when users face problems with the chat agent in ScramCorp app? NOT FOR: Use listUseCases for topic/cluster analysis of conversations. Use listAiAgents to get agent IDs and names first when the user has not specified an agent. RETURNS: - Table of issues: summary, instance count, conversation count, agentId, appId, visitorIds, and accountIds. - Sorted by conversation count descending. Limited to 500 rows. The maximum time range for this tool is 90 days.

Pendo data is split into subscriptions, which share a set of visitors and accounts. Each subscription is split into separate applications. This call returns a list of all the names and ids of all the subscriptions this user has access to, along with the names and ids of all of the applications in that subscription. Most tools for this mcp require at least a subscription id; many also require an application id. Application ids are not unique across different subscriptions.

List a subscription's designated business OBJECTS - the custom event properties that have been marked as analyzable business entities (e.g. 'dashboardId', 'venueId', 'orderId'). Returns each object's underlying event property name (its field) and kind, which are exactly the objectProperty argument the objectAnalytics* tools take. Objects are NOT pages, features, or track events. Returns property names (e.g. 'dashboardId'), not human display labels. USE FOR: Discovering which business objects a subscription has, and grounding an ambiguous entity name before an objectAnalytics* call. Use it when a name could mean either a business object or a page/feature (e.g. 'dashboards') to confirm the object exists and get the exact objectProperty.field to pass to objectAnalyticsActiveCount or objectAnalyticsBreakdown. EXAMPLES: - What business objects can I analyze in this subscription? - Is 'dashboard' a business object or a page? - List the custom objects available for Object Analytics - Which object property identifies venues? NOT FOR: Listing pages, features, or track events - use listCountables. Counting or ranking objects, or any per-object metric - use objectAnalyticsActiveCount or objectAnalyticsBreakdown. This returns only the object property definitions, never metrics. RETURNS: - One entry per designated object property: field (the event property name to pass as objectProperty.field, e.g. 'dashboardId'), and kind ('event' or 'historical'). - For historical (promoted visitor/account/parentAccount metadata) objects, group and metadataKind are also included.

List the guide delivery order ("throttle order") for an app. When multiple guides are eligible to show at the same time, the delivery order determines which guide takes precedence. This tool returns the ordered list of guides for the given app. An app with no ordering set returns an empty list. Guides are capped by limit (default 50). totalGuides is the full ordering length; offset and returned describe the slice in guides. When returned < totalGuides the result is truncated - tell the user which range they're seeing (e.g. "showing 1-50 of 120") and use offset to page through the rest. PRESENTATION (follow exactly): For EVERY guide, show ALL SIX of these columns, always in this order: Guide Name | Status | Segment | Page | Guide Category | Product Area Show every column even when most guides share a value and even when a column has no values. Segment is the readable segment name (e.g. "Everyone", "Browser: Chrome"). Page is "Sitewide" when the guide targets no specific page. Do NOT show the raw guideId to the user. It is for your context only, to reference guides in follow-up tool calls. EXAMPLES: - What is the guide delivery order for app xxx? - Which guide shows first when several are eligible? - Show me the throttle ordering for app xxx RETURNS: - appId, appName, totalGuides, offset, returned, and the ordered guides for the app - totalGuides/offset/returned: the full ordering length and the returned slice - surface the range to the user when truncated - guides: in delivery order. Always display these six columns per guide: Guide Name, Status, Segment, Page, Guide Category, Product Area - guideId is context-only - never shown to the user

List all product areas in the subscription with their IDs, names, and descriptions.

Returns a list of themes for a subscription. Themes define the visual styling applied to guides and other in-app content. Supports optional filtering by application and fuzzy search. USE FOR: Listing available themes, finding a theme by name, or getting a theme ID to reference in other tools. Use when the user asks about available themes or visual styles for their guides. EXAMPLES: - What themes do we have? - List all themes for this subscription - Show me themes for app 12345 - Find a theme named 'Dark Mode' NOT FOR: Modifying or creating themes. Viewing guide content or guide metrics. RETURNS: - Array of themes with: id, name, appId, tags, archive status, buildingBlocks (visual styling properties), cssUrl (supplemental CSS file URL if present)

Returns tracked (curated) use case definitions and their associated conversation and event IDs for a given AI agent and time window. Tracked use cases are user-defined categories; conversations are attributed to them by LLM classification. USE FOR: Listing all tracked use cases defined for an agent and discovering which conversations belong to each. Returns conversationIds and eventIds that identify the conversations attributed to each tracked use case for further analysis. EXAMPLES: - What tracked use cases are defined for my agent? - Which conversations belong to the 'billing support' tracked use case? - Show me the tracked use cases for my agent over the last 30 days. NOT FOR: Unsupervised clustering of conversations into emergent topics, or aggregate volume metrics - this tool only lists predefined tracked use cases and the conversations attributed to them. RETURNS: - Per tracked use case: id, name, summary, numConversations, conversationIds, eventIds. The maximum time range for this tool is 90 days.

List the visitors that match a segment and get a summary of the cohort. Returns {summary, rows}: summary has numVisitors and numAccounts (the true segment totals, independent of limit); rows is the list of matched visitors with visitorId and any requested metadata fields, capped by limit. The cohort is defined by a segmentPipeline - either a saved Pendo segment reference or a full inline pipeline produced by the segment-builder tool. USE FOR: Listing the members of a segment/cohort and getting its visitor and account totals. EXAMPLES: - List the visitors in segment X - Who are the visitors in this cohort? - How many visitors and accounts are in this segment? - Show me visitors in segment X with their email and role RETURNS: - summary: numVisitors, numAccounts (segment totals, independent of limit) - rows: visitorId plus requested fields in a metadata object keyed by metadata path (for example, "metadata.auto.lastvisit"), capped by limit

Analyze the individual business OBJECTS (e.g. dashboards, venues, documents, orders) of one kind over a date range, where an object is identified by one event property. Use this for questions about a business object - including when a page or feature shares the same name (e.g. treat 'dashboards' as objects, not the 'Dashboards' page). An object is a custom event property designated as an analyzable entity, distinct from pages, features, and track events. Two modes, selected by the 'measures' argument: (1) RANKING (default, measures omitted or ['uniqueVisitors']): a ranked list of objects by how many unique visitors were active on each - 'which <objects> had the most/least visitors'. Returns one row per object (objectId, objectName, uniqueVisitors), top-N by sortBy, optionally scoped by a segment. (2) AVERAGES (measures includes 'avgObjectsPerVisitor' and/or 'avgTimePerObject'): a single subscription-wide summary scalar per measure - 'on average how many <objects> does each visitor use', 'average time spent per <object>' - matching the object-details page tiles. These are aggregates across ALL objects, NOT per-object rows, so sortBy/limit do not apply. The aggregation is built by the Pendo analytics gateway, not in this service. USE FOR: Ranking the individual objects of one business-object kind by distinct active visitors - e.g. which dashboards had the most visitors, busiest venues, most-visited documents - even if a page or feature shares the object's name. Also the home for per-object AVERAGE and TIME questions about a business object, e.g. 'average time per dashboard' or 'average dashboards per visitor'. When the object name is ambiguous (could be a page/feature) or you don't know its event property, ground it with listCustomObjects first. EXAMPLES: - Which dashboards had the most visitors last month? - Which venues had the most unique visitors in the last 30 days? - Top 10 documents by distinct visitors between 2025-01-01 and 2025-01-31 - Least-visited order IDs over the last week - On average, how many distinct dashboards does each visitor interact with over the last 30 days? - On average, how much time do visitors spend on each dashboard over the last 30 days? NOT FOR: A single 'how many unique objects' scalar - use objectAnalyticsActiveCount. Per-page, per-feature, or per-track-event usage - use aggregateEntityUsage. Per-visitor or per-account app usage - use appUsage. Page dwell-time or time-on-page - that is page analytics, not a business object; 'time spent on a business object' is this tool, page dwell-time is not. RETURNS: - meta.dateRange: resolved startDate and endDate (YYYY-MM-DD); meta.compareToDateRange when comparing; meta.measures: the requested measures - meta.objectProperty: the field and kind analyzed - RANKING mode: rows - one per object with objectId, objectName, uniqueVisitors, ranked by sortBy. In comparison mode uniqueVisitors splits into current_/prior_/change_/pctChange_ columns. - AVERAGES mode: summary - a single object keyed by the requested average measure(s), e.g. {avgObjectsPerVisitor, avgTimePerObject}; no rows.

Rank or trend the events/actions on one kind of business OBJECT (e.g. dashboards, venues, documents, orders) over a date range - the event types (pages, features, track events) fired while the object's identifying property is present. Two modes: (1) default - rank those events by count for 'what are the most common actions/events on a <object>'; (2) set trend='up' or 'down' - rank which events are TRENDING up or down on the object, comparing the requested window against the equal-length window immediately before it (rows carry currentPeriodEvents, previousPeriodEvents, pctChange, ranked by pctChange). This is THE tool for 'which events are trending up/down across my <objects>' - it scopes to the object and computes the period-over-period change for you; do not hand-compute a trend by pulling two windows of unscoped track events from another tool. An object is a custom event property designated as an analyzable entity (distinct from pages, features, and track events), including when a page or feature shares the object's name. Returns one row per event type. The aggregation is built by the Pendo analytics gateway, not in this service. USE FOR: The events/actions on a business object - either the most common ('the most common actions taken on dashboards this month', 'top interactions on a document') or, with trend='up'/'down', which are TRENDING ('which events are trending up across dashboards', 'what actions on venues are trending down vs the previous period'). When the object name is ambiguous (could be a page/feature) or you don't know its event property, ground it with listCustomObjects first. EXAMPLES: - What are the most common actions taken on dashboards this month? - Top 10 event types on venues in the last 30 days - Which actions were performed most on documents between 2025-01-01 and 2025-01-31? - Which events are trending up across dashboards in the last 30 days? - What actions on venues are trending down compared with the previous period? NOT FOR: How many unique objects were active - use objectAnalyticsActiveCount. Ranking the objects themselves by visitors (which dashboards had the most visitors) or per-object averages/time - use objectAnalyticsBreakdown. Per-page, per-feature, or per-track-event usage NOT scoped to a business object - use aggregateEntityUsage. Do NOT answer 'which events are trending on <objects>' by comparing two windows of aggregateEntityUsage/entityUsage yourself - that counts ALL events unscoped to the object; set trend on this tool instead. RETURNS: - meta.dateRange: resolved startDate and endDate (YYYY-MM-DD) - meta.objectProperty: the field and kind analyzed - meta.trend: the trending direction, when trend mode is requested - meta.compareToDateRange: the previous equal-length window (startDate, endDate) the trend compares against, when trend mode is requested - rows (default): one per event type performed on the object, each with eventId, eventName, eventKind (page/feature/track type) and numEvents, ranked by numEvents - rows (trend mode): one per event type, each with eventId, eventName, eventKind, currentPeriodEvents, previousPeriodEvents and pctChange, ranked by pctChange

Calculate the Product Engagement Score for an application over a date range, returning adoption, stickiness, and growth metrics.

Search for product entities such as pages, features, track types, guides, accounts, and segments.

Return the set of metadata fields available for visitors.