DecisionBox
VerifiedAutonomous agent that explores warehouses and delivers verified insights.
What is DecisionBox?
DecisionBox is an open-source autonomous agent that connects directly to your data warehouse and independently decides which areas to investigate. It generates and runs dozens of SQL queries per session, validates results through independent checks, and produces severity-ranked insights with actionable recommendations.
The system works by first linking to warehouses such as BigQuery or Snowflake, selecting a domain configuration, then launching discovery runs. During execution the agent iterates on findings in real time while recording every query, reasoning step, and decision for full auditability.
It is designed for data-driven teams that need reliable, hands-off analysis without writing prompts or manually exploring tables. Users can review outputs, provide feedback, and refine future runs through semantic search or direct questions.
What you can build with DecisionBox
E-commerce performance review
Run periodic discovery to identify underperforming product categories and inventory issues without manual queries.
Gaming metrics monitoring
Let the agent surface player retention patterns and monetization opportunities across large event tables.
Social platform health checks
Automatically detect engagement drops or content trends and receive ranked next-step recommendations.
Install DecisionBox
# Clone the repository
git clone https://github.com/decisionbox-io/decisionbox-platform.git
cd decisionbox-platform
# Start MongoDB + Qdrant + API + Dashboard
docker compose up -d
# Open the dashboard
open http://localhost:3000- 1Clone the DecisionBox repository and build the Go binary.
- 2Connect your data warehouse credentials through the configuration file.
- 3Select or create a domain pack and choose an LLM provider.
- 4Launch a discovery run and monitor progress in the dashboard.
- 5Review validated insights and act on the numbered recommendations.
DecisionBox: pros & cons
Pros
- +Fully autonomous exploration without user prompts
- +Every insight is independently re-queried for accuracy
- +Complete logging of SQL, reasoning, and decisions
- +Self-hostable with support for local models
Cons
- –Requires direct warehouse access and permissions
- –Discovery runs can incur LLM and query costs
- –Domain customization needs separate pack authoring
Frequently asked questions
It works with BigQuery, Redshift, Snowflake, PostgreSQL, Databricks, and Microsoft SQL Server.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other general-purpose options worth comparing.
