Common Amplitude MCP tasks
You describe the goal in plain language and your client picks the tools. The tool names in the Tools it uses column tell you which specific tools are running. This lets you verify results or ask for a specific tool.
Look up users and their activity
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Find a specific user's recent events | "What has user 12345 done in the last 7 days?" | get_users, query_amplitude_data |
| Look up a user's profile and properties | "Show the user properties for user 12345." | get_users |
| Compare two groups of users | "Compare converters and drop-offs in the onboarding flow." | query_amplitude_data, create_cohort |
Inspect events and properties
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Understand an event's properties and the values they take | "List the properties on Checkout Started and the values plan_type takes." | get_events, get_properties, query_amplitude_data |
| Find an event or property by name | "Which events mention 'checkout'?" | search, get_events, get_properties |
| Review custom events and transformations | "Show the custom events that merge signup events." | get_custom_or_labeled_events, get_transformations, get_group_types |
| Document or rename taxonomy | "Set the description of plan_type to 'Subscription tier'." | update_properties, update_event, update_custom_or_labeled_events |
Analyze data
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Answer an ad-hoc question (segmentation, funnel, retention) | "Show signup conversion by traffic source this month." | query_amplitude_data, render_chart |
| Read an existing chart or dashboard | "Summarize this dashboard for my exec meeting. [dashboard URL]" | get_from_url, get_dashboard, get_charts, query_charts |
| Investigate a spike or drop | "Why did weekly active users drop last Tuesday?" | query_amplitude_data, get_deployments, get_experiments |
| Analyze an experiment | "Should we ship the checkout experiment? Analyze the results." | get_experiments, query_experiment |
Watch session replays
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Find replays for a behavior or page | "Show replays where users rage-clicked on the checkout page." | get_session_replays, list_session_replays |
| Reproduce a reported bug | "Find this user's session and walk me through what happened." | get_session_replays, get_session_replay_events |
Create and edit content
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Build a chart, dashboard, or notebook | "Build an executive dashboard for our product launch." | create_dashboard, create_notebook, render_chart, save_chart_edits |
| Edit an existing dashboard or notebook | "Add a retention chart to the KPI dashboard." | edit_dashboard, edit_notebook |
| Create an experiment or feature flag | "Create a 50/50 experiment for the new checkout flow." | create_experiment, create_flags, create_metric, update_flag |
| Build a cohort and sync it | "Create a power-users cohort and sync it to Iterable." | create_cohort, get_cohort_sync_destinations, sync_cohort |
Manage tracking-plan branches
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Make taxonomy changes on a branch | "Create a branch q3-cleanup and add a Checkout Started event." | get_branches, create_branch, create_events, create_properties |
| Ship or discard branch changes | "Merge q3-cleanup into main." | refresh_branch, merge_branch, delete_branch |
Review customer feedback
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Summarize what customers are asking for | "What are the top customer complaints this month?" | get_feedback_insights, get_feedback_mentions |
| Track a theme over time | "How has feedback about the mobile app trended over the last 90 days?" | get_feedback_trends, get_feedback_sources |
Analyze AI agent performance
| Goal | Ask something like | Tools it uses |
|---|---|---|
| Measure agent quality, cost, and errors | "How are our AI agents performing? Show quality, cost, and error rates." | query_agent_analytics_metrics, query_agent_analytics_sessions |
| Investigate a bad agent session | "Investigate the low-quality AI sessions from yesterday." | query_agent_analytics_spans, get_agent_analytics_conversation, search_agent_analytics_conversations |
Check data warehouse connections and jobs
| Goal | Ask something like | Tools it uses |
|---|---|---|
| List import sources for a project | "What warehouse imports and other data sources are configured for this project?" | get_data_ingestion_sources |
| Check one source's configuration | "Show me the configuration for my Snowflake import." | get_data_source_details |
| List export destinations | "Where does this project export its data?" | get_data_warehouse_destinations |
| Check recent import or export jobs | "Show me the recent jobs for my Snowflake source. Which ones failed or completed with errors?" | get_data_ingestion_sources or get_data_warehouse_destinations, then get_data_warehouse_jobs |
| Review manual backfill exports | "Show me the manual backfill export jobs for my Snowflake destination." | get_data_warehouse_jobs |
When to use something else
The Amplitude MCP server reads and edits Amplitude content with your account permissions. Some jobs belong to a different surface:
- Connect your own external tools or a third-party MCP server to an Amplitude agent: use Custom Agents and Agent Connectors, not the Amplitude MCP server.
- Chat with AI about your data inside Amplitude, with no client setup: use the Global Agent or the Dashboard Agent.
- Send event data: use an Amplitude SDK or the HTTP V2 API. The MCP server isn't an ingestion endpoint.
- Read Amplitude's public documentation from an AI client: use the read-only Docs MCP server.
Not sure a tool exists or what it needs? In your client, ask "What Amplitude MCP tools are available for [job]?" or turn on progressive tool discovery to browse tool categories and inspect each tool's parameters before you call it.
Was this helpful?