W
About
Vanna AI is a production-ready, open-source SQL agent framework designed to make data accessible to everyone in an organization. By leveraging large language models, Vanna converts natural language questions into SQL queries across any supported database—including PostgreSQL, MySQL, Snowflake, BigQuery, and SQLite. It works with leading LLM providers such as OpenAI, Anthropic, Gemini, and Ollama, offering true flexibility with no vendor lock-in. The framework is built for real-world deployment, featuring multi-turn conversational context, access control, and multi-database support out of the box. Vanna's optional cloud layer adds enterprise-grade capabilities including user-scoped permissions, real-time observability with distributed tracing, hosted vector database memory for continuous learning, secure file storage, comprehensive audit logs, and long-term data retention for compliance. Developers can get started with a single `pip install vanna` and connect to a dummy database in minutes. Vanna is ideal for data teams, product managers, and business analysts who need fast, accurate answers from complex data systems without deep SQL expertise. It ships to production without rewrites, scales with quota management, and integrates rich UI components directly into the agent experience.
Key Features
- Natural Language to SQL: Translates plain English questions into accurate, multi-turn SQL queries, enabling non-technical users to query any database without writing code.
- Multi-LLM & Multi-Database Support: Works with OpenAI, Anthropic, Gemini, and Ollama, and connects to PostgreSQL, MySQL, Snowflake, BigQuery, SQLite, and more databases.
- Access Control & User Permissions: Built-in user-scoped execution and access control ensure each user only sees the data they're authorized to view, making it enterprise-safe.
- Observability & Audit Logs: Real-time metrics, distributed tracing, and comprehensive audit logs provide full visibility into agent behavior and compliance tracking.
- Hosted Vector DB Memory: Agent memory via a hosted vector database allows Vanna to continuously learn from past queries and improve accuracy over time.
Use Cases
- Enabling business analysts and product managers to query company databases in plain English without SQL expertise
- Embedding a natural language data interface into SaaS products with per-user access control and quota management
- Accelerating data exploration for engineering teams by reducing time spent writing repetitive SQL queries
- Building internal data Q&A tools that learn from historical queries using vector database memory
- Deploying a compliant, audit-logged SQL agent in enterprise environments with multi-database access
Pros
- Truly Open Source: The core framework is fully open source with no vendor lock-in, allowing teams to self-host or use any LLM and database combination.
- Production-Ready from Day One: Ships with user permissions, quota management, and rich UI components baked in—no rewrites required when scaling to enterprise.
- Fast to Get Started: A single pip install and a few minutes of setup get you querying a database in natural language, with a dummy database available for immediate testing.
Cons
- Advanced Features Require Cloud Plan: Enterprise capabilities like hosted vector memory, audit logs, and observability are only available through the paid Vanna Cloud offering.
- Developer Setup Required: Initial configuration and integration with existing databases and LLM providers requires Python development skills and environment setup.
Frequently Asked Questions
What databases does Vanna AI support?
Vanna supports SQLite, PostgreSQL, MySQL, Snowflake, BigQuery, and many more databases. See the documentation for the full list of supported connectors.
Which LLM providers can I use with Vanna?
Vanna works with OpenAI, Anthropic (Claude), Google Gemini, and Ollama for local models, giving you full flexibility over your AI stack.
Is Vanna AI free to use?
Yes, the Vanna core framework is open source and free. Optional hosted admin features like access control, observability, and audit logs are available on paid Cloud plans.
Can non-technical users use Vanna AI?
Absolutely. Once deployed by a developer, business users can query databases using plain English questions without any SQL knowledge.
How do I get started with Vanna AI?
Run `pip install vanna` to install the package, then follow the quickstart guide in the documentation. A preconfigured dummy database in Colab lets you test in under a minute with no sign-in required.
