Vanna

Overview

Vanna AI is an open-source, production-ready SQL agent framework designed to accelerate access to enterprise data. It allows users to query databases using natural language, automatically translating questions into accurate SQL statements. Vanna positions itself as “the fast SQL Agent” and is built to handle complex, multi-turn interactions across an organization's systems of record. It is not merely a text-to-SQL wrapper; it is a complete framework for deploying AI agents that execute queries, return results, and integrate with production workflows. The platform works with any major LLM provider, giving teams the flexibility to select the model that best fits their accuracy, latency, and cost requirements.

Vanna supports multiple LLM providers including Anthropic, OpenAI, Gemini, and Ollama, and connects to common databases such as SQLite, PostgreSQL, MySQL, Snowflake, BigQuery, and others via its extensible architecture. Multi-turn conversation enables users to iterate on their questions and refine results without needing deep SQL expertise. The open-source core ensures full customization and no vendor lock-in, while the optional hosted admin layer provides enterprise-grade control. Hosted features include access control to restrict data visibility, observability with real-time metrics and distributed tracing, agent memory backed by a hosted vector database, a secure file system for storage and retrieval, comprehensive audit logs for compliance, and long-term data retention for conversations and logs. These features are available through Vanna Cloud or the Enterprise edition.

Getting started can be as simple as using a dummy database in Google Colab in about one minute with no sign-in required. For self-hosting, teams install the framework using pip install vanna and then configure their preferred LLM and database connection. The framework is designed for production readiness from the ground up, supporting both lightweight internal tools and enterprise-scale deployments. Vanna Cloud provides a managed alternative for organizations that want to avoid infrastructure overhead. The combination of open-source portability and optional cloud management makes Vanna suitable for startups, mid-sized companies, and large enterprises.

The primary value of Vanna is reducing the time spent writing SQL and accelerating the generation of insights. By enabling natural-language interaction, it democratizes access to data for non-technical employees, product managers, and executives. Engineering teams can embed Vanna as a backend for custom data applications, while data teams can use it to automate reporting and ad-hoc analysis. The platform also enables governed access to sensitive systems of record, with audit trails and access control meeting security and compliance needs. Vanna ultimately empowers organizations to build AI agents their users can actually use, transforming how data is accessed, queried, and acted upon.

Core capabilities

Multi-database support

Connect to SQLite, PostgreSQL, MySQL, Snowflake, BigQuery, and many other database systems.

Any LLM provider

Integrate with Anthropic, OpenAI, Gemini, Ollama, or any other model provider via the open framework.

Multi-turn conversation

Iteratively refine natural-language queries to zero in on the exact data and answer you need.

Access control

Hosted admin features let you control which users or roles can see which data objects.

Observability

Built-in real-time metrics and distributed tracing help monitor agent behavior and query performance.

Agent memory

A hosted vector database stores conversation context and learned preferences for better answers over time.

Secure file system

Store and retrieve files securely to support richer agent workflows and data retrieval.

Audit logs

Track all user activities and queries for compliance, security, and governance requirements.

Data retention

Keep long-term records of conversations and logs to support analysis and regulatory needs.

Use cases

Natural-language querying of enterprise databases for ad-hoc insights

Enabling non-technical employees to explore data without SQL knowledge

Building internal SQL agents for customer support or operations teams

Automating routine reporting workflows with AI-generated SQL

Querying multiple database systems through a unified conversational interface

Providing governed, auditable access to sensitive data for compliance-conscious organizations

Accelerating data discovery and reducing backlog for data engineering teams