DecisionBox is an open-source autonomous AI discovery agent designed to connect to data warehouses and databases. The system runs autonomous AI agents that write and execute SQL queries, then surfaces validated insights and actionable recommendations without requiring users to ask questions or write queries.
The platform supports multiple data warehouse providers including BigQuery, Redshift, Snowflake, Databricks, and PostgreSQL. For PostgreSQL specifically, it includes safety features such as read-only access by Postgres role, TLS encryption by default, capped connection pools, per-query timeouts, and no write permissions. The system operates entirely through read-only access with no DDL or DML capabilities.
DecisionBox works by connecting to databases and running autonomous discovery loops. Users connect the agent to their database, point it at domain packs, and the system writes its own SQL queries, validates findings against the data, and produces a ranked backlog of insights and recommendations. The PostgreSQL integration supports both field-by-field configuration and full DSN connections, making it compatible with managed PostgreSQL services including RDS, Supabase, Neon, Railway, Fly.io, Aiven, Crunchy, and others.
The system is designed for production environments with safety measures including capped connection pools, per-query timeouts, and read-only access enforced at the database role level. All code for the PostgreSQL provider is available in the public repository under AGPL v3 license. The same autonomous agent runs across all supported database platforms, allowing organizations to maintain consistent discovery capabilities as their data infrastructure evolves.
DecisionBox is suitable for organizations using PostgreSQL in production environments who need autonomous data discovery capabilities. The platform is particularly valuable for teams managing live traffic databases who require insights without impacting performance. The open-source nature and read-only design make it appropriate for security-conscious environments.
Key Features
- •The agent independently decides what to investigate based on your data and domain pack, writing and executing SQL queries in an autonomous loop. It performs 50–100+ queries per run without any prompting, surfacing hidden insights and anomalies.
- •Connect DecisionBox to BigQuery, Redshift, Snowflake, Oracle, SAP HANA, and more with automatic schema discovery. This ensures consistent autonomous discovery across your entire data infrastructure without manual configuration.