# Mixpanel MCP Prompt Library — Agent Reference > Optimized for LLMs and AI agents. Use as a context file (Claude project, Cursor, etc.) or reference loaded by an MCP skill. ## How to use this library Match user intent to a prompt below. Replace `[brackets]` with specifics. Resolve `~~category` placeholders to whatever tool is connected (see Connectors). When a request spans multiple prompts, compose them into a single workflow — do not ask the user to run them separately. ## Rules 1. **Start with schema discovery.** Call Get-Events and Get-Event-Details to confirm event names before any analysis. Do not guess. 2. **Every analysis prompt needs four things:** behavior (events), population (who), timeframe (when), answer shape (rate, trend, breakdown). 3. **Verify properties before breakdowns.** Call Get-Property-Names and Get-Property-Values first. Wrong names fail silently. 4. **Flows for discovery, Funnels for measurement.** Use `chartType=sankey` for Flows, especially with `stepsBefore`. 5. **Prefer property filters over cohorts.** Cohorts are supported, but `plan_type = enterprise` is more transparent than a named cohort. 6. **Check data quality before critical analysis.** Call Get-Issues on unvalidated events. Broken events produce confident wrong answers. 7. **Cross-tool prompts require connected MCPs.** Check what's available before suggesting. Only offer chains where a connector is active. 8. **Writes require Admin or Project Owner.** If a write returns an error or empty response, inform the user about the permission requirement. Do not retry. 9. **Rate limit: 600 requests/hour.** Break 20+ query sessions into multiple sessions. 10. **Always use explicit dates.** Anchor all timeframes to today's date in ISO format. Do not rely on "last 30 days" without computing the actual range. 11. **Set the project first.** If Get-Projects returns multiple and the user hasn't specified one, ask before running queries. Do not default. ## Anti-patterns Never do these: - **Don't run a funnel without confirming event names.** The query will succeed with zero results and look like a real finding. - **Don't assume a property exists because it sounds right.** Always verify with Get-Property-Names. - **Don't filter by cohort name as a string.** Cohort IDs are required for cohort-based filters. - **Don't interpret zero results as "this doesn't happen."** Check the event name, property values, and date range first. - **Don't present a single data point as a trend.** If you only have one period, say so — don't frame it as directional. ## Confidence signals Flag uncertainty in these situations: - Query returns zero results → verify event name and filters before interpreting. - Retention or conversion looks unusually high or low → call out sample size. - AI-generated severity or prioritization → note it's inference, not a score from the API. - Flows return fewer paths than requested → note the data may be sparse for this timeframe. --- ## Connectors Cross-tool prompts use `~~category` as a placeholder. Resolve to the connected tool at runtime. Before suggesting a cross-tool workflow, check which MCP servers are available. If detection isn't possible, ask once at session start: "Which other tools are connected alongside Mixpanel?" | Placeholder | Chain with Mixpanel | Examples | |---|---|---| | `~~chat` | Search feedback, team threads about features/bugs | Slack, Teams | | `~~calendar` | Meeting prep — pull usage for upcoming accounts | Google Calendar, Outlook | | `~~project tracker` | Match drop-offs to bugs/tasks, file issues | Jira, Linear, Asana | | `~~error monitoring` | Correlate error events with exceptions | Sentry, Datadog, Bugsnag | | `~~knowledge base` | Export data dictionaries, write findings | Notion, Confluence | | `~~CRM` | Enrich accounts with deal stage, ARR, owner | Salesforce, HubSpot | | `~~product feedback` | Match usage with feature requests, complaints | Productboard, Canny, Zendesk | | `~~dev environment` | Bring usage data into the building workflow | Cursor, Lovable, Replit, Windsurf | | `~~enterprise search` | Find related internal docs, past analyses | Glean, Kendra | | `~~email` | Draft account updates, share summaries | Gmail, Outlook | | `~~data warehouse` | Join behavioral data with revenue/billing | BigQuery, Snowflake, Databricks | | `~~ai observability` | Correlate LLM performance with user behavior | Langfuse, Helicone, Braintrust | --- ## Prompts Fields: `section`, `category`, `title`, `intent` (what the user is trying to do), `prompt`, `tools`, `response_format` (how to present results), and optional `caveat`, `follow_up`, `chain_with`, `connectors`, `tags`. ```yaml # ============================================================ # ORIENT — Schema discovery. Run these first. # ============================================================ - section: orient title: Find your projects intent: User needs to identify which Mixpanel project to work with prompt: "What Mixpanel projects do I have access to? List them with their IDs." tools: [Get-Projects] response_format: Table of project name and ID tags: [setup, project-id] - section: orient title: Map schema to a business question intent: User wants to know which events represent a behavior they care about prompt: "I want to understand [onboarding completion / feature adoption / checkout conversion] in this project. Look at the available events and properties and tell me which ones best represent that behavior. Suggest the top 3–5 candidates and explain your reasoning." tools: [Get-Events, Get-Event-Details] response_format: Ranked list of candidate events with rationale caveat: Quality depends on Lexicon descriptions. follow_up: "Give me the full details on [Event Name], including any data quality issues." tags: [schema, events, discovery] - section: orient title: Understand what a specific event means intent: User wants full context on a single event prompt: "Give me the full details on the [Event Name] event: its description, what properties are attached, when it fires, and any data quality issues. Tell me if anything looks off." tools: [Get-Event-Details, Get-Issues, Get-Property-Names] response_format: Event summary with properties list and any quality flags tags: [schema, event-details, data-quality] - section: orient title: Find your activation signal intent: User wants to identify the best activation event for their product prompt: "Based on the events in this project, what's the best candidate for an activation event — something that predicts whether a new user will stick around? Show me the most relevant events and explain why each one does or doesn't fit." tools: [Get-Events, Get-Event-Details] response_format: Recommended activation event with reasoning caveat: Reasoning task only — cannot validate statistically. Follow up with retention query. chain_with: [N-day retention from a starting event] tags: [activation, discovery, reasoning] - section: orient title: Explore properties and their actual values intent: User needs to know what filters and breakdowns are available prompt: "What properties are available on the [Event Name] event? For the ones useful as filters or breakdowns (plan type, device, user role), show me the actual values each one contains." tools: [Get-Property-Names, Get-Property-Values, Get-Property] response_format: Property list with sample values for each caveat: Run this before any breakdown prompt. tags: [schema, properties, filters, breakdowns] - section: orient title: Get a direct link to Lexicon intent: User wants to edit an event in the Mixpanel UI prompt: "Give me the Lexicon URL for the [Event Name] event so I can go edit it in the Mixpanel UI." tools: [Get-Lexicon-URL] response_format: Direct URL tags: [lexicon, link] # ============================================================ # ANALYZE — Funnels, retention, flows, adoption, trends. # ============================================================ # -- Funnels -- - section: analyze category: funnels title: Conversion funnel with a specific population intent: User wants to measure conversion through a series of steps prompt: "What's the conversion rate from [Step 1] to [Step 2] to [Step 3] for [first-time users / paid accounts / mobile users] over the last 30 days? Show me where the biggest drop-off is." tools: [Get-Query-Schema, Run-Query] response_format: Step-by-step conversion with drop-off percentages. Lead with the biggest drop. caveat: Run Orient prompts first if unsure which property to filter on. follow_up: "Break that down by [plan type / device / channel]." chain_with: [Discover what happens between two funnel steps, Replay after identifying users via analysis] tags: [funnels, conversion, population, breakdown] - section: analyze category: funnels title: Week-over-week funnel trend intent: User wants to see if conversion is improving or declining prompt: "Show me how our conversion rate from [Event A] to [Event B] has changed week over week for the past 8 weeks. Flag any weeks where conversion dropped more than 10%." tools: [Get-Query-Schema, Run-Query] response_format: Weekly conversion rates with flagged drops caveat: Flagging is AI reasoning, not a tool feature. tags: [funnels, trends, week-over-week] - section: analyze category: funnels title: Funnel breakdown by segment intent: User wants to compare conversion across segments prompt: "Break down our [Event A] to [Event B] conversion by [plan type / acquisition channel / device type] for the last 60 days. Which segment converts best, which converts worst, and what's the gap?" tools: [Get-Query-Schema, Run-Query] response_format: Segments ranked by conversion rate with gap analysis caveat: Verify the property name exists before running. tags: [funnels, breakdown, segments] # -- Retention -- - section: analyze category: retention title: N-day retention from a starting event intent: User wants to measure how many users come back after a key action prompt: "What's the 1, 7, 14, and 30-day retention for users who completed [Starting Event] in the last 90 days? Use [Return Event] as the retention signal." tools: [Get-Query-Schema, Run-Query] response_format: Retention curve — D1, D7, D14, D30 percentages follow_up: "Compare that to users who completed [Alternative Starting Event]." tags: [retention, n-day] - section: analyze category: retention title: Retention comparison across two time periods intent: User wants to know if retention is getting better or worse prompt: "Compare 30-day retention for users who signed up in [Month 1] versus [Month 2]. Did retention improve? Which part of the curve changed: early drop-off or long-term engagement?" tools: [Get-Query-Schema, Run-Query] response_format: Side-by-side retention curves with delta analysis caveat: Runs two queries and compares. Needs sufficient users in both periods. tags: [retention, comparison, time-periods] # -- Feature Adoption -- - section: analyze category: feature-adoption title: Adoption rate for a specific feature intent: User wants to know what percentage of their users have tried a feature prompt: "What percentage of [active users / paid accounts / users who completed onboarding] have used [Feature Event] at least once in the last 30 days? How does that compare to the 30 days before?" tools: [Get-Query-Schema, Run-Query] response_format: Adoption percentage, current vs. prior period with delta tags: [adoption, feature, period-comparison] - section: analyze category: feature-adoption title: Feature usage depth intent: User wants to know how heavily a feature is used, not just whether it's tried prompt: "For users who have used [Feature Event] at least once, how many times do they use it on average per week? Break it down by [plan type / user role] if those properties exist." tools: [Get-Property-Names, Get-Query-Schema, Run-Query] response_format: Average frequency per segment caveat: Verify breakdown property exists first. tags: [adoption, frequency, breakdown] # -- Trends -- - section: analyze category: trends title: Event volume trend intent: User wants to spot anomalies in event volume prompt: "Show me the daily volume of [Event Name] over the last 30 days. Flag any days with a spike or drop greater than 20% from the 7-day average." tools: [Get-Query-Schema, Run-Query] response_format: Daily counts with flagged anomalies and magnitude caveat: Spike detection is AI reasoning. For persistent monitoring, recreate in Mixpanel with an alert. tags: [trends, volume, anomaly-detection] - section: analyze category: trends title: Pull a saved report intent: User wants to retrieve an existing report by name prompt: "Retrieve the report named [Report Name] and show me the current results." tools: [Get-Report] response_format: Report results summarized with headline finding caveat: Use Search-Entities to find reports by name if needed. tags: [reports, saved, retrieval] - section: analyze category: trends title: Comparative period analysis intent: User wants a high-level health check across multiple metrics prompt: "Compare this month to last month across these three metrics: [DAU], [Signup-to-Purchase conversion], and [7-day retention]. Summarize what improved, what got worse, and what stayed flat." tools: [Run-Query] response_format: Three metrics, each with current/prior/delta. Lead with what changed most. caveat: 6-query session. Keep each query simple — add breakdowns as a follow-up. follow_up: "The conversion rate dropped. Break that funnel down by [channel / device] to see which segment drove the change." tags: [trends, comparison, multi-metric] # -- Flows -- - section: analyze category: flows title: What do users do after a key event? intent: User wants to discover what happens next after a specific action prompt: "Show me the 3 most common steps users take after [Event Name] in the last 30 days. Where do they go, and how many drop off at each step?" tools: [Get-Query-Schema, Run-Query] response_format: Top paths with step counts and drop-off at each node caveat: Use stepsAfter=3, stepsBefore=0, chartType=sankey. follow_up: "Now show me the same thing but only for users where [plan_type = free]." tags: [flows, paths, post-event, sankey] - section: analyze category: flows title: What leads to a conversion event? intent: User wants to understand the paths that precede a key conversion prompt: "What are the 3 most common paths users take in the steps before [Purchase Completed / Subscription Started / Activation Event] in the last 30 days? Which paths are most common?" tools: [Get-Query-Schema, Run-Query] response_format: Pre-conversion paths ranked by volume caveat: Use stepsBefore=3, stepsAfter=0, chartType=sankey. Paths chart type may return empty with stepsBefore. tags: [flows, paths, pre-event, conversion, sankey] - section: analyze category: flows title: Compare paths between two segments intent: User wants to see where different user groups diverge in behavior prompt: "Compare the top user paths after [Onboarding Completed] for [free users] versus [paid users] in the last 30 days. Where do their journeys diverge?" tools: [Get-Query-Schema, Run-Query] response_format: Side-by-side dominant paths with divergence points highlighted caveat: Runs two queries with different filters and compares. tags: [flows, segments, comparison] - section: analyze category: flows title: Discover what happens between two funnel steps intent: User has a funnel drop-off and wants to understand what users do instead prompt: "I have a funnel from [Step A] to [Step B] where conversion is low. Show me the most common paths users take between those two events. What are they doing instead of converting?" tools: [Get-Query-Schema, Run-Query] response_format: Intermediate paths with counts, highlighting which lead to conversion vs. drop-off tags: [flows, funnels, diagnostic, drop-off] # -- Feature Launch -- - section: analyze category: feature-launch title: End-of-sprint feature assessment intent: User shipped a feature and wants a full evaluation prompt: "We shipped [Feature Name] two weeks ago. Run these analyses in order: 1) weekly adoption trend (unique users of [Feature Event] per week for the last 4 weeks), 2) a funnel from [first use] to [repeat use], 3) retention for users who adopted vs. users who didn't, and 4) the top paths users take after [Feature Event]. Summarize whether the launch is on track." tools: [Run-Query] response_format: Four-part summary — adoption trend, repeat conversion, retention delta, path analysis. End with go/no-go assessment. caveat: 4–7 tool call session. chain_with: [Replay for a specific user, Create a dashboard from scratch] tags: [feature-launch, adoption, funnels, retention, flows, multi-step] # ============================================================ # INVESTIGATE — User-level and session-level analysis. # ============================================================ - section: investigate title: Replay for a specific user intent: User has a user ID and wants to see what they did prompt: "Pull session replays for user [distinct_id] from the last 14 days. I want to see what they were doing around the time they [dropped off / stopped using the feature / triggered an error]." tools: [Get-User-Replays-Data] response_format: Replay links with session timestamps and summary of key actions caveat: Works best with a user ID from prior analysis or a support ticket. tags: [replay, user, session] - section: investigate title: Replay after identifying users via analysis intent: User wants to find drop-off users and watch their sessions prompt: "Step 1: Show me the user IDs of users who reached [Step N] but did not complete [Final Event] in the last 7 days. Return up to 10 distinct_ids. Step 2: Now pull session replays for those users." tools: [Get-Query-Schema, Run-Query, Get-User-Replays-Data] response_format: User IDs with replay links. Note which step each user reached. caveat: MCP returns replay metadata and links. Recordings open in Mixpanel UI. tags: [replay, funnels, drop-off, two-step] - section: investigate title: Diagnose a drop-off user intent: User wants to understand why a specific person stopped using the product prompt: "User [distinct_id] was active last month but hasn't logged in for 2 weeks. Pull their last 30 days of activity and any session replays from their final sessions. What were they doing before they stopped?" tools: [Run-Query, Get-User-Replays-Data] response_format: Activity timeline showing usage arc, with replay links for final sessions tags: [replay, churn, user, drop-off, activity] - section: investigate title: Review a user's recent sessions intent: User wants to see whether engagement is getting deeper or shallower prompt: "Pull the last 5 sessions for user [distinct_id]. For each session, show me what features they used, how long the session lasted, and where they exited. Are their sessions getting shorter or less engaged over time?" tools: [Get-User-Replays-Data, Run-Query] response_format: Session-by-session breakdown with trajectory assessment follow_up: "Show me the replay for their shortest session." tags: [replay, sessions, engagement, trajectory] - section: investigate title: Account-level investigation intent: User wants a usage overview for a specific company or account prompt: "Show me all activity for users at [Account Name / company_id] over the last 30 days. Which features are they using most? Where are they dropping off?" tools: [Get-Events, Run-Query] response_format: Feature usage ranked by volume, with drop-off points flagged caveat: Requires company properties on events (e.g., company_id). follow_up: "Pull session replays for their most active user." tags: [account, investigation, features, engagement] # ============================================================ # BUILD — Dashboards and project artifact management. # ============================================================ - section: build category: dashboards title: Create a dashboard from scratch intent: User wants a persistent dashboard with specific reports prompt: "Create a Mixpanel dashboard called [Dashboard Name] with these reports: 1) daily active users over the last 30 days, 2) signup-to-purchase conversion funnel, 3) 7-day retention for new users. Add a text card at the top summarizing what this board tracks." tools: [Create-Dashboard] response_format: Confirm dashboard created with link caveat: Dashboards persist in Mixpanel after the conversation ends. follow_up: "Add a fourth report showing feature adoption by plan type." tags: [dashboard, create, reports, persist] - section: build category: dashboards title: Build a weekly growth dashboard intent: User wants an ongoing growth tracking board prompt: "Build a dashboard that tracks signups, activations, and churn week over week for the last 12 weeks. Include a text card at the top explaining the metric definitions." tools: [Create-Dashboard] response_format: Confirm dashboard created with link caveat: Discover event names via Orient prompts first if unsure. tags: [dashboard, growth, weekly] - section: build category: dashboards title: Duplicate and customize an existing dashboard intent: User wants a copy of a dashboard with modifications prompt: "Duplicate the [Dashboard Name] dashboard and change the date range to last quarter. Rename it [New Dashboard Name]." tools: [Get-Dashboard, Duplicate-Dashboard] response_format: Confirm new dashboard with link tags: [dashboard, duplicate, customize] - section: build category: dashboards title: Update a dashboard's layout intent: User wants to rearrange or add to an existing dashboard prompt: "In the [Dashboard Name] dashboard, move the retention report to the top row and add a new text card below it explaining how we define retention." tools: [Get-Dashboard, Update-Dashboard] response_format: Confirm layout updated caveat: Call Get-Dashboard with include_layout=True first to get cell/row IDs. tags: [dashboard, layout, update] - section: build category: audit title: List and audit existing dashboards intent: User wants to find and clean up stale dashboards prompt: "What dashboards exist in this project? Which ones haven't been updated in 90 days?" tools: [Search-Entities] response_format: Dashboard list with last-edited dates, stale ones flagged caveat: Search-Entities also finds experiments, flags, metric trees, playlists, heat maps, cohorts — but can only drill into dashboards and reports. follow_up: "Delete the ones that are stale. Actually, show me their contents first." tags: [audit, dashboards, stale] - section: build category: audit title: Full project artifact inventory intent: User wants to understand everything that exists in the project prompt: "Give me an inventory of everything in this project: dashboards, reports, cohorts, experiments, feature flags, and metric trees. For each type, tell me the count and when the most recent one was last edited." tools: [Search-Entities] response_format: Entity types with counts and last-edited dates caveat: Metadata only for experiments, flags, and metric trees — can't inspect configuration. tags: [audit, inventory, artifacts, hygiene] - section: build category: audit title: Find stale experiments and feature flags intent: User wants to identify experiments and flags that need cleanup prompt: "List all experiments and feature flags in this project. Which ones haven't been edited in the last 60 days? Who created them?" tools: [Search-Entities] response_format: List with creator, last-edited date, stale flag caveat: No access to variants, targeting rules, or results. tags: [audit, experiments, feature-flags, stale] # ============================================================ # GOVERN — Lexicon management and data quality. # ============================================================ - section: govern category: lexicon title: Add descriptions to undocumented events intent: User wants to fill in missing Lexicon descriptions at scale prompt: "Find all events that don't have a description in Lexicon. For each one, suggest a description based on the event name and its properties. Then apply the descriptions." tools: [Get-Events, Get-Event-Details, Edit-Event] response_format: List of events with suggested descriptions, confirm before applying caveat: Review before confirming — AI can guess wrong with internal shorthand. tags: [lexicon, descriptions, governance, write] - section: govern category: lexicon title: Tag related events intent: User wants to organize events by category prompt: "Tag all events related to [checkout / onboarding / search] with the tag \"[Tag Name]\". Create the tag if it doesn't exist yet." tools: [Get-Events, Create-Tag, Edit-Event] response_format: Count of events tagged, list of names follow_up: "Now do the same for all properties on those events." tags: [lexicon, tags, governance, write] - section: govern category: lexicon title: Hide inactive events intent: User wants to clean up unused events from the UI prompt: "Find all events that haven't fired in the last 90 days and hide them in Lexicon." tools: [Get-Events, Edit-Event] response_format: Count hidden, list of event names caveat: Hidden events are still queryable by name. tags: [lexicon, hide, cleanup, write] - section: govern category: lexicon title: Flag PII properties intent: User wants to audit properties for personally identifiable information prompt: "Look at all properties across all events. Flag any that look like they might contain personally identifiable information (email, phone, name, IP address) but aren't marked as sensitive yet." tools: [Get-Events, Get-Property-Names, Get-Property-Values, Edit-Property] response_format: List of flagged properties with reasoning. Starting audit, not final pass. tags: [lexicon, pii, sensitive, governance, write] - section: govern category: lexicon title: Rename a tag across the project intent: User wants to rename a tag globally prompt: "Rename the tag \"[Old Tag]\" to \"[New Tag]\" across all events and properties that use it." tools: [Rename-Tag] response_format: Confirm rename with count of affected entities tags: [lexicon, tags, rename, write] - section: govern category: data-quality title: Audit data quality for a specific event intent: User wants to know if a specific event has problems prompt: "Are there any open data quality issues for the [Event Name] event? Summarize what's broken, when each issue was first detected, and how long it's been flagged. Tell me which ones to fix first." tools: [Get-Issues] response_format: Issues ranked by age with recommended fix order caveat: Prioritization is inference, not an API severity score. tags: [data-quality, audit, event-specific] - section: govern category: data-quality title: Full project health check intent: User wants a broad view of data quality across the project prompt: "Run a data quality audit on this project. What are the most critical open issues across all events and properties? Group them by severity and tell me which ones are most likely affecting analysis right now." tools: [Get-Issues] response_format: Issues grouped by inferred severity, most impactful first tags: [data-quality, audit, project-wide] - section: govern category: data-quality title: Dismiss resolved issues in bulk intent: User wants to clean up resolved data quality issues prompt: "Dismiss all data quality issues for events we deprecated last quarter. Specifically, dismiss issues for [Event A], [Event B], and [Event C]." tools: [Dismiss-Issues] response_format: Confirm dismissed with count tags: [data-quality, dismiss, cleanup, write] - section: govern category: data-quality title: Check data quality before critical analysis intent: User is about to run an important analysis and wants to verify the data first prompt: "Before I run a funnel analysis on [Event A] through [Event C], check for any data quality issues on those three events. Tell me if anything would make the results unreliable." tools: [Get-Issues] response_format: Clean/not-clean verdict per event, with details on any issues found chain_with: [Conversion funnel with a specific population] tags: [data-quality, pre-check, funnels] - section: govern category: data-quality title: Escalate stale data quality issues intent: User wants a formatted list of old issues to send to their data team prompt: "Find all data quality issues that have been open for more than 14 days. For each one, give me the event name, issue description, and how long it's been flagged. Format it so I can paste it into a message to the data engineering team." tools: [Get-Issues] response_format: Copy-pasteable list formatted for chat or issue tracker tags: [data-quality, escalation, stale, formatting] - section: govern category: data-quality title: Data quality trend check intent: User wants to know if data quality is improving or degrading prompt: "How many open data quality issues does this project have right now? Compare that to the issues created in the last 30 days versus the last 60 days. Are we accumulating issues faster than we're resolving them?" tools: [Get-Issues] response_format: Issue counts by window with directional trend caveat: Directional only — dismissed issues drop from the count. tags: [data-quality, trend, accumulation] # ============================================================ # CHAIN — Cross-tool and file-based workflows. # ============================================================ # -- Attached Files -- - section: chain category: attached-files title: Correlate a metric change with external events intent: User sees a metric shift and wants to understand what caused it prompt: "Our [metric] changed between [date range]. I'm attaching our [campaign calendar / release notes / experiment log]. Does the timing correlate with anything in this document? What's the most likely explanation?" tools: [] response_format: Timeline alignment showing which external events coincide with the metric change caveat: No MCP tools — AI reasons across query results and attached file. tags: [synthesis, file-attachment, correlation] - section: chain category: attached-files title: QBR prep with template intent: User is preparing a quarterly business review for an account prompt: "I'm preparing a QBR for [Account Name]. I've attached our slide template. Pull their usage data for the last quarter, then fill in the template with: feature adoption rates, engagement trends, and any areas of concern." tools: [Run-Query] response_format: Filled template sections with data caveat: Requires file attachment support. tags: [qbr, template, account, file-attachment] - section: chain category: attached-files title: Strategic synthesis intent: User wants to combine analytics findings with external context for recommendations prompt: "Based on the analysis results above, plus the [benchmark doc / customer interview notes / competitive intel] attached, what are the two or three highest-priority changes for [the onboarding flow / this feature / this segment's experience] going into next quarter? Be direct." tools: [] response_format: Prioritized recommendations with supporting evidence from both data and context caveat: No MCP tools — synthesis only. Best after running analysis prompts. tags: [synthesis, strategy, file-attachment] # -- Cross-Tool -- - section: chain category: cross-tool title: Correlate usage data with customer feedback intent: User wants to combine quantitative usage with qualitative sentiment prompt: "Pull adoption numbers for [Feature X] this week from Mixpanel, then search ~~chat for any mentions of [Feature X] in customer-facing channels. Summarize the quantitative data and qualitative feedback together." tools: [Run-Query] connectors: [chat] response_format: Usage metrics + feedback themes, with assessment of whether sentiment matches the numbers caveat: Search is keyword-based — add alternate names if customers call the feature something different. tags: [cross-tool, chat, feedback, adoption] - section: chain category: cross-tool title: Meeting prep with usage data intent: User has upcoming meetings and wants usage context for each account prompt: "Check ~~calendar for external meetings this week. For each company, pull their usage data from Mixpanel for the last 30 days. Summarize each account in 3 bullets with anything that changed week-over-week." tools: [Run-Query] connectors: [calendar] response_format: Per-account bullets with usage trends caveat: Project needs a company/account property on events. tags: [cross-tool, calendar, meetings, account, prep] - section: chain category: cross-tool title: Match drop-offs to open bugs or tasks intent: User wants to check if known issues explain funnel drop-offs prompt: "Show me the top 3 drop-off points in our [Core Funnel] from Mixpanel. Then check ~~project tracker for any open issues or tasks related to those drop-off points." tools: [Run-Query] connectors: [project_tracker] response_format: Drop-off points paired with matching issues (or noted as untracked) caveat: Results depend on how the team titles issues. No matches is useful information. tags: [cross-tool, project-tracker, bugs, funnels] - section: chain category: cross-tool title: Correlate errors with user impact intent: User wants to connect backend errors with frontend behavioral impact prompt: "Pull the users who triggered [Error Event] more than 3 times this week in Mixpanel. Then check ~~error monitoring for exceptions from those same users or matching error messages. Are these the same root cause or different bugs?" tools: [Run-Query] connectors: [error_monitoring] response_format: User overlap analysis with root cause assessment caveat: If different user identifiers across tools, specify how to match (email, distinct_id). tags: [cross-tool, error-monitoring, debugging] - section: chain category: cross-tool title: Export schema to team wiki intent: User wants their Mixpanel schema documented in an accessible format prompt: "Pull the schema for this project — all events, their descriptions, and their properties. Then create a page in ~~knowledge base with a formatted data dictionary." tools: [Get-Events, Get-Event-Details] connectors: [knowledge_base] response_format: Formatted data dictionary page tags: [cross-tool, knowledge-base, data-dictionary, schema] - section: chain category: cross-tool title: Enrich account investigation with CRM data intent: User wants usage data combined with deal context prompt: "Pull usage data for [Account Name] from Mixpanel for the last 30 days. Then check ~~CRM for their deal stage, ARR, renewal date, and account owner. Combine into a single account summary." tools: [Run-Query] connectors: [crm] response_format: Unified account brief — usage metrics + deal context caveat: Requires company properties on Mixpanel events and matching account in CRM. tags: [cross-tool, crm, account, enrichment] - section: chain category: cross-tool title: Match usage patterns with feature requests intent: User wants to connect behavioral data with what customers are asking for prompt: "Pull the top features by usage volume this month from Mixpanel. Then search ~~product feedback for feature requests or complaints related to those features. Where does usage data disagree with customer sentiment?" tools: [Run-Query] connectors: [product_feedback] response_format: Features ranked by usage with matched feedback. Highlight mismatches. tags: [cross-tool, product-feedback, feature-requests, sentiment] - section: chain category: cross-tool title: Bring usage data into a dev workflow intent: Developer wants analytics context while building prompt: "Pull the adoption and drop-off data for [Feature Area] from Mixpanel. Summarize which parts users engage with most and where they struggle. I'm building in ~~dev environment and need this context to prioritize what to fix." tools: [Run-Query] connectors: [dev_environment] response_format: Concise usage summary optimized for dev context — what works, what doesn't, where to focus tags: [cross-tool, dev-environment, building, prioritization] - section: chain category: cross-tool title: Find related internal context intent: User wants to find prior analyses, docs, or discussions related to their Mixpanel findings prompt: "I just found that [metric or finding] in Mixpanel. Search ~~enterprise search for any related internal documents, past analyses, or strategy memos that might provide context." tools: [] connectors: [enterprise_search] response_format: List of related documents with relevance summary tags: [cross-tool, enterprise-search, context, discovery] - section: chain category: cross-tool title: Draft and share analysis summary intent: User wants to send findings to stakeholders prompt: "Summarize the analysis we just completed in 5 bullet points. Draft an ~~email to [recipient] with the subject '[Topic] — Usage Update' sharing the key findings and recommended next steps." tools: [] connectors: [email] response_format: Email draft ready to send tags: [cross-tool, email, sharing, summary] - section: chain category: cross-tool title: Join behavioral data with warehouse metrics intent: User wants to combine Mixpanel behavioral data with revenue or backend data prompt: "Pull feature adoption data for [Feature X] from Mixpanel. Then query ~~data warehouse for revenue per user or account over the same period. Do users who adopt [Feature X] generate more revenue?" tools: [Run-Query] connectors: [data_warehouse] response_format: Adoption vs. revenue correlation with sample sizes caveat: Requires matching user/account identifiers across Mixpanel and the warehouse. tags: [cross-tool, data-warehouse, revenue, correlation] - section: chain category: cross-tool title: Correlate LLM performance with user behavior intent: User building an AI product wants to connect model performance with UX outcomes prompt: "Pull user engagement metrics for [AI Feature] from Mixpanel — usage frequency, session depth, and any error events. Then check ~~ai observability for latency, error rates, and model performance over the same period. Are model issues correlated with user drop-off?" tools: [Run-Query] connectors: [ai_observability] response_format: Side-by-side performance and behavioral metrics with correlation assessment tags: [cross-tool, ai-observability, llm, performance, correlation] ```