Mitzu
mitzu.io
- Free trial
- Web
- API
- Self-hosted
What Mitzu is
Mitzu is an agentic product-analytics platform that queries event data in a customer’s data warehouse. Users can ask funnel, retention, and cohort questions in plain language through the app or Slack, with additional analysis for segmentation, user journeys, and B2B accounts. A Configuration Agent scans warehouse tables to build a semantic layer that analysts can review and adjust. Its deterministic engine generates SQL from analysis specifications, and users can inspect that SQL. Supported warehouses include Snowflake, BigQuery, Databricks, Redshift, ClickHouse, and AWS Athena; recognized schemas include Segment, Snowplow, Firebase, GA4, RudderStack, and custom events. Mitzu describes a zero-copy architecture, monitoring for metric shifts and retention or activation anomalies, and email or Slack alerts. Enterprise deployment can use private VPC or self-hosting. Analyst costs $134 a month billed yearly, with a 14-day trial and no free plan.
Who it suits
Mitzu suits teams whose event data already lands in a supported warehouse and analysts who want natural-language questions with visible SQL. Organizations needing private VPC or self-hosted deployment can use its enterprise options.
Good to know, and what to watch
- Natural-language funnel, retention, and cohort queries
- Visible SQL generated from analysis specifications
- Zero-copy warehouse querying
- Metric anomaly alerts through email or Slack
- Private VPC and self-hosted enterprise deployment
- Requires event data in a warehouse
- No free plan
- Analyst plan limits usage to three Editor seats
Plans
Every plan and priceEnterprise
Analyst (yearly)
Analyst
Team (yearly)
14-day trial
The verdict
Mitzu connects warehouse event data with natural-language product analysis, semantic modeling, and monitoring. Its warehouse requirement and paid, seat-based plans are key fit considerations.
What the maker publishes
- AI privacy
- The security page says only query context is sent to AI, customer data is not used to train or fine-tune LLMs, and metadata logs can be deleted upon request. source
- Analytics
- The platform supports funnels, retention analysis, segmentation, cohorts, user journeys, and B2B account analytics. source
- CDP and event schemas
- Mitzu says it recognizes Segment, Snowplow, Firebase, GA4, RudderStack, and custom event schemas in warehouse data. source
- Company
- The site identifies the company as Mitzu, Inc. and lists István Mészáros as co-founder and CEO. source
- Data handling
- Mitzu describes its architecture as zero-copy: queries run against warehouse data without copying or replicating raw event data. source
- Deployment
- Enterprise customers can use private VPC or self-hosted deployment, including within their own VPC or on-premise environment. source
- Limits and intended fit
- Mitzu says it requires event data to be landing in a warehouse, and that it may not fit organizations whose events are locked in a tool that will not export them. source
- Monitoring
- Mitzu says it monitors warehouse metrics and surfaces metric shifts, retention anomalies, and activation drops, with alerts via email or Slack. source
- Natural language
- Users can ask funnel, retention, and cohort questions in plain language through the app or Slack. source
- Product
- Mitzu is an agentic product analytics platform that queries event data in a customer’s data warehouse. source
- Semantic layer
- A Configuration Agent scans warehouse tables and builds a semantic layer that analysts can review and adjust. source
- SQL visibility
- Mitzu says its deterministic query engine generates SQL from analysis specifications, and users can inspect the SQL behind queries. source
- Support
- Analyst includes email support, Team includes priority email support, and Enterprise lists dedicated Slack support, SLA services, onboarding, and training. source
- Warehouse integrations
- Mitzu names Snowflake, BigQuery, Databricks, Redshift, and ClickHouse as supported warehouses; its security page also lists AWS Athena. source
Checked against the maker's pages on 9 Oct 2026.