Umami 3.4 launches with MCP support, annotations, account API keys, and a typed API client

Open-source web analytics platform Umami has unveiled version 3.4, marking a significant step forward in bridging traditional analytics workflows with AI-powered tools. The release introduces Model Context Protocol (MCP) support, enabling compatible AI assistants to directly query analytics data through natural language questions about traffic, funnels, revenue, and performance metrics. Developers can connect via remote MCP endpoints or run the @umami/mcp package locally, with all tools operating in read-only mode and respecting existing user permission checks.

The update also brings contextual annotations to analytics dashboards, allowing users to add dated notes marking product launches, campaigns, pricing changes, or outages directly alongside traffic charts. This creates a shared historical record for teams, eliminating reliance on institutional memory when investigating anomalies. For self-hosted deployments, new account API keys enable authentication for scripts and integrations without browser login tokens, while a newly released TypeScript API client (@umami/api-client) provides typed access to organized endpoints across funnels, goals, performance reporting, and UTM metrics.

Additional improvements include session property filters in saved segments, a new data-distinct-id tracker attribute for visitor identity management, and expanded language support with Azerbaijani and Georgian translations. Performance gains focus on PostgreSQL query optimization for visitor counts and session activity, while mobile experience refinements address date pickers, two-factor authentication code pasting, and sidebar navigation. The release maintains backward compatibility with existing report API routes, allowing gradual migration to feature-specific endpoints without disrupting established integrations.