Automate Amplitude Data Management

A skill for automating Amplitude product analytics through Composio's Amplitude toolkit via Rube MCP. It documents five core workflows with their tool sequences, parameters, and pitfalls: sending events (AMPLITUDE_SEND_EVENTS, requiring event_type plus either user_id or device_id, with time in 13-digit millisecond epoch, subject to a per-request batch limit); getting user activity (AMPLITUDE_FIND_USER to resolve an application user_id to Amplitude's internal ID, then AMPLITUDE_GET_USER_ACTIVITY, paginated via offset/limit and returned in reverse chronological order); finding and identifying users (AMPLITUDE_FIND_USER, which searches across user_id, device_id, and Amplitude ID, plus AMPLITUDE_IDENTIFY using Amplitude's $set/$setOnce/$add/$append/$unset property operations); managing cohorts (AMPLITUDE_LIST_COHORTS, AMPLITUDE_GET_COHORT, AMPLITUDE_UPDATE_COHORT_MEMBERSHIP with add/remove user-ID arrays, and AMPLITUDE_CHECK_COHORT_STATUS polled with the request_id returned from the update call, since only behavioral cohorts support API membership updates); and browsing event categories via AMPLITUDE_GET_EVENT_CATEGORIES to validate event_type values before sending events, since category names are case-sensitive. It also documents common patterns for resolving an application user_id to Amplitude's internal ID before calling GET_USER_ACTIVITY, resolving a cohort name to a cohort_id via LIST_COHORTS, and polling CHECK_COHORT_STATUS until an async cohort update completes or errors.

Use it when a user wants to send events to Amplitude, look up a user's event history, identify or update user properties, manage cohort membership, or discover available event categories. Rube MCP must be connected with an active Amplitude connection (verified via RUBE_MANAGE_CONNECTIONS, completing the returned auth link if the connection isn't ACTIVE), and RUBE_SEARCH_TOOLS should always be called first to get current tool schemas before running any workflow.

Inputs are workflow-specific: event arrays with event_type/user_id or device_id/event_properties/user_properties/time for sending events, a search term for user lookup, or a cohort_id and membership add/remove arrays for cohort operations. Outputs include Amplitude's internal user ID, user event streams in reverse chronological order, cohort lists and details (which may include archived cohorts, distinguished by a status field), and request_id values for polling async cohort operations. Responses may be nested under a data key and should be parsed defensively with fallbacks for optional fields.

Connects to Amplitude exclusively through Composio's amplitude toolkit via Rube MCP (https://rube.app/mcp, no API keys needed client-side) - no direct Amplitude API key handling is required. Known pitfalls include event timestamps requiring millisecond precision (10-digit seconds values are misread as very old dates), per-project event-ingestion rate limits that favor batching events where possible, and user-property changes and cohort membership updates both being eventually consistent rather than immediate.

Developers and analysts automating Amplitude workflows - sending tracking events, auditing individual user activity, managing user properties, or maintaining cohort membership - through an AI assistant connected via Rube MCP instead of hand-rolling direct Amplitude API calls and their own ID-resolution and pagination logic.