nao

nao is an open source analytics agent builder for context engineering. It is described as an analytics agent built for context engineering, and it is aimed at building, evaluating, and deploying analytics agents on a data stack. The product page also says it can deploy a chat UI for anyone to run analytics on data.

Its central idea is to build agent context like a file system. The context can include data, metadata, rules, documentation, tools, and MCPs, and the page states there is no limit on what can be added. nao can also synchronize existing context automatically from sources such as databases, repositories, and external sources. The examples shown include BigQuery-style database configuration, a dbt repository, and Notion pages.

The same interface supports testing and deployment. nao includes a command to test context reliability by creating unit tests of questions to SQL and returning metrics such as test pass rate, answer rate, average time, and total tokens. It also offers a chat UI for asking questions in plain English, with examples of data exploration and data stories. The page says it can be deployed with a user’s own LLM key, with token consumption billed only for usage, and it can be asked from platforms including Slack, Teams, WhatsApp, Telegram, Claude, Codex, and Cursor through nao MCP. It also supports self-hosting, with the claim that everything stays in the user’s infrastructure under their control.

The product is marked as 100% open source. The page also includes links and navigation for pricing, docs, resources, GitHub, integrations, security, changelog, blog, events, and legal pages, and it identifies Nao Labs in the footer.