TL;DR
TL;DR: Nao Labs (YC X25) forks VS Code with built-in BigQuery, Snowflake, and Postgres connectors, feeding your full data schema into an AI copilot that writes SQL, Python, and YAML with real context — not blind autocompletion.
Source and Accuracy Notes
- HN Launch: Launch HN: Nao Labs (YC X25) – Cursor for Data (158 points)
- Official site: getnao.io
- Downloads: sunshine.getnao.io/releases
- Demo: YouTube walkthrough
What Is Nao Labs?
Cursor and Copilot transformed general-purpose coding. But when you work with data — writing SQL pipelines, managing dbt models, debugging a dashboard that suddenly shows wrong revenue figures — generic AI assistants keep falling short. They autocomplete SQL they can’t verify, suggest code without knowing your schema, and give you no visibility into what a change will do to your actual data output.
Nao Labs, a YC X25 batch company, built the tool that gap. Nao is a fork of VS Code purpose-built for data teams. It ships with native connectors to BigQuery, Snowflake, and Postgres, maintains a live RAG of your entire warehouse schema, and has a dedicated AI copilot that writes SQL, Python, and YAML with full schema context baked in.
The core idea: when you ask an AI to write code that touches your data, you care about the output — not the code itself. Nao keeps that front and center by showing you code diffs and data diffs side by side, so you can see exactly what a change will do before it reaches production.
How Nao Works
AI Copilot With Schema Context
Unlike a generic AI assistant that sees only the code in front of it, Nao maintains a RAG of your warehouse schemas — tables, columns, relationships, dbt models — always available to the copilot. When you start typing a SQL query, the AI knows your user_events table has a session_id column that joins to sessions.id. It won’t suggest joining on a column that doesn’t exist.
The copilot also tracks your data lineage. Before you commit a change, it can assess downstream impact — which dashboards, models, or pipelines will be affected if this column gets renamed or this aggregation gets changed.
Data Diffs and Quality Checks
Nao shows code diffs and data diffs side by side. Run your pipeline locally against dev data and Nao shows you exactly what changed in the output rows. This is a direct answer to the anxiety of “will this multiply revenue by 5x on the CEO dashboard?” — you see the answer before shipping.
The built-in agent also runs automated quality checks: detects missing or duplicated values, flags outliers, compares dev vs. production data, and alerts when a change will break downstream models.
Data Warehouse Connectors
Nao ships with first-party connectors for:
- BigQuery
- Snowflake
- Postgres
Authentication is handled through the editor — no need to juggle credentials across a warehouse console, a BI tool, and your IDE separately.
dbt Integration
Most data teams managing SQL pipelines use dbt. Nao understands dbt models, documentation, and tests natively. The AI agent can help create new models, update documentation, run tests, and assess lineage impact across your dbt project — all from within the editor.
VS Code Fork
Since Nao is a VS Code fork, your existing extensions, keybindings, and workflows mostly carry over. Git works as expected. Terminal is there. The learning curve is minimal for anyone already in the VS Code ecosystem.
Why Not Just Use Cursor + MCPs?
The natural question: Cursor already has an AI copilot and supports MCPs for data sources. Why not just connect Snowflake or BigQuery MCP to Cursor?
The Nao founders argue this doesn’t deliver the same experience. MCPs in Cursor are “very enclosed” — they don’t bring data context into every tab, they require manual setup of authentication and pipelines, and they don’t adapt the UI for data workflows. You still end up context-switching.
Nao’s approach is pre-packaged: install, authenticate to your warehouse, and the full schema is available everywhere in the editor. Non-technical data team members can use it without configuring MCP infrastructure.
Practical Evaluation Checklist
Strengths:
- Full schema context in every AI interaction — no more blind SQL autocompletion
- Side-by-side code and data diffs for pre-ship verification
- Native dbt integration for pipeline authors
- Automated data quality checks built into the workflow
- VS Code fork means low switching cost for existing users
- Free to start — no credit card required
Limitations:
- Mac only at launch (Linux and Windows coming)
- Currently supports BigQuery, Snowflake, and Postgres — no Redshift, Databricks, or ClickHouse yet
- Heavy data warehouse context means the RAG needs to be kept in sync as schemas evolve
- As a VS Code fork, updates may lag upstream VS Code releases
Ideal use cases:
- Data engineers writing and maintaining SQL pipelines, especially with dbt
- Analytics engineers doing schema modeling and quality assurance
- Data analysts who want AI-assisted SQL without writing blind
- Teams where data quality bugs in production are costly
Security Notes
Nao authenticates directly with your data warehouse. The AI processing is local to the editor — your credentials don’t flow through an external API by default. For organizations with strict data residency requirements, this local-first authentication model is preferable to sending query results to an external LLM API. Always review your warehouse’s access controls and ensure Nao’s service account has only the minimum required permissions.
FAQ
Q: Does Nao work with dbt Core or only dbt Cloud? A: The HN announcement mentions dbt integration for pipeline management, documentation, and testing. Specific support scope (Core vs. Cloud) should be verified on the official docs at getnao.io.
Q: Is my data sent to OpenAI or Anthropic for AI processing? A: Nao has its own AI copilot. The architecture details around which models power the copilot and where data is processed are on their site — check the privacy policy and architecture docs for specifics if data residency is a concern.
Q: Can I use Nao without connecting a warehouse first? A: The editor itself functions as a VS Code fork, but the core value (schema-aware AI, data diffs) requires connecting at least one supported warehouse. Temporary accounts are available for trying it out without committing credentials.
Q: Does Nao work with existing VS Code extensions? A: Most VS Code extensions should work since Nao is a fork. That said, extensions that directly access the filesystem or interact with specific VS Code APIs may have compatibility edge cases worth testing.
Q: What’s the pricing? A: Free to start. Specific pricing tiers are on the getnao.io website.
Conclusion
The “Cursor for X” framing gets used a lot, but Nao actually earns it in a meaningful way. The problem it solves — AI that writes SQL without knowing your schema, and code changes you ship without seeing their data impact — is real and painful for data teams. Nao’s approach of maintaining a live schema RAG, surfacing data diffs before commit, and integrating natively with dbt addresses those pain points directly rather than bolting an MCP onto a general-purpose editor.
If your team lives in SQL, manages a data warehouse, and wants AI assistance that actually understands your data, Nao is worth an afternoon to evaluate. The Mac-only limitation is a blocker for mixed-OS teams today, but the dbt integration and schema-aware copilot make it a promising direction for data engineering workflows.
- Official site: getnao.io
- Downloads: sunshine.getnao.io/releases