nao

nao is an open-source analytics agent builder. Data teams use it to build the context an analytics agent needs, test the agent’s answers, and deploy a chat interface so anyone can ask questions in plain language. It is licensed Apache 2.0, so you can self-host it or run it on nao’s cloud.

What it does

nao has two parts. The first is a command-line tool that data teams use to build context. You connect a warehouse, sync the schema, define metrics, write down business rules, and add tests. The second is a chat interface where business users ask questions and get answers, deployed on the web or inside Slack, Teams, Telegram, or WhatsApp.

The idea behind it is context engineering. nao treats the agent’s context like code that you build, version, and test, rather than something you hope the model figures out.

How it works

A data team runs the nao CLI to set up the context, then checks the agent’s accuracy with tests before anyone relies on it. The agent answers questions in a loop, writing SQL, running it on the warehouse, checking the result, and correcting itself. Because the context is owned by the team, a wrong answer can be traced to a missing rule or definition and fixed.

What it connects to

nao reads from common warehouses including BigQuery, Snowflake, Postgres, Databricks, Redshift, and DuckDB. It works with dbt for models, and it can pull knowledge from sources like Notion, Confluence, and Google Drive. It runs on several model providers, including Claude, GPT, and Gemini.

Who it is for

nao fits data-led companies where wrong numbers are costly and a data team can own the context. The buyer is usually a head of data or a founding engineer. It is a strong fit if you want control, testing, and governance. It is a weaker fit if you want a zero-setup tool and you are fine trading control for speed, since nao expects you to build the context first.

Strengths and trade-offs

The strengths are control, openness, and a real testing workflow, which is what makes the answers trustworthy. The trade-off is setup. nao asks a data team to invest in the context up front. That work is the source of the reliability, so it is a deliberate cost rather than a flaw, but it does mean nao is not the right pick for a team with no data function.

To see how nao compares to other approaches, read the best agentic analytics tools.