Skip to content

ZLens

Certified answers on your warehouse.

ZLens is an open-source analytics workbench. It works on top of your dbt project, or straight on your database's own tables. You build blocks (reviewed, reusable queries), turn them into Apps, and Ask questions in plain English. Every answer and every tile says how far to trust it: Certified, Governed, Needs review, or a refusal with the reason.

Using dbt? ZLens reads your models, tests and lineage. Not using dbt? Start straight from your database.

Everything is a file in your project, so it is reviewed in git like code. ZLens runs on your machine; no hosted account is needed.

ZLens in under a minute Ask a question and get an answer from a certified block, labelled Certified answer; see how it was answered; open the App built from the same certified blocks, where all data is certified; then trace lineage from the dbt sources and models through the block to the App.
  1. Ask "How has revenue changed by month?".
  2. The answer is labelled Certified answer, from the monthly_revenue block, with a table and a line chart.
  3. How it was answered shows the steps Ask took.
  4. Open the Jaffle Growth Review App; it reads All data certified and every tile is labelled Certified.
  5. Open Lineage and trace monthly_revenue; the graph runs from dbt sources and models to the block, its dashboard and the App.

Start here

  1. Quickstart: run ZLens on a sample dbt project and get your first result in about ten minutes. No dbt? Start from your database.
  2. Ask a question and read its trust label.
  3. Certify a block so Ask and Apps can trust it.

Follow the journey
Get started Install, open the workbench, connect a database and an AI provider, add ZLens to your dbt project or start from your database alone.
Ask and trust Certified answers, answers that need review and their saved drafts, refusals with a reason, and how each answer was made.
Build blocks Import SQL, work in Block Studio, save drafts and certify them.
Notebooks SQL and DQL cells, results, saving a cell as a block, and sharing through git.
Apps and dashboards Read an App and its trust labels, build and publish one, ask questions about it.
Lineage From dbt models to blocks to Apps, in the workbench and the terminal.
Connect your AI agent Use ZLens from Claude Code, Codex, Cursor or Claude Desktop over MCP.
Reference CLI, DQL language, file formats, project layout, connectors.

How trust works

  • Saving never certifies. Anything saved, imported or kept from an AI answer is a draft until a person clicks Certify.
  • Ask tries the most trusted source first: a certified block, then the semantic layer, then SQL written by AI, which is marked for review and saved as a draft.
  • When ZLens cannot answer safely, it says so, and why, instead of guessing.

Packages

Package What it is
@zlens/cli The zlens command and the workbench
create-zlens-app Creates a new ZLens project, on a dbt project or a database alone
@zlens/mcp The MCP server for AI agents

GitHub · Roadmap · Support