Legible Documentation
Legible is an open-source GenBI (Generative Business Intelligence) agent. Ask questions in natural language and get SQL queries, charts, and insights from your data.
What is Legible?
Legible connects to your existing data sources (PostgreSQL, BigQuery, DuckDB, Snowflake, and more) and uses AI to translate natural language questions into accurate SQL. It includes a semantic modeling layer (MDL) that understands your data relationships, so the AI generates correct, context-aware queries.
Key Features
- Natural Language to SQL — Ask questions in plain English and get SQL + results
- 22+ Data Source Connectors — PostgreSQL, BigQuery, Snowflake, DuckDB, MySQL, ClickHouse, and more
- Semantic Modeling (MDL) — Define models, relationships, and calculated fields for smarter AI queries
- MCP Server — Connect AI agents (Claude, Cursor, Cline) directly to your data via the Model Context Protocol
- Self-Hosted — Full control over your data and infrastructure
- Extensible — Supports custom LLMs, embedding models, and vector stores
Quick Links
- Quickstart — Get running in 5 minutes
- Architecture — Understand how the components fit together
- Configuration — Customize LLMs, embeddings, and more
- MCP Integration — Connect AI agents to your data
- Docker Deployment — Production Docker setup
- Kubernetes Deployment — K8s with Kustomize