Databench by Alkera
Self-hostable open-source workspace where your team works together with AI agents on notebooks, SQL queries and shared files.
What it does
Databench is a web-based workspace for data work that you run on your own infrastructure. In a notebook you mix SQL, Python and agent cells; query results arrive in Python directly as a dataframe. AI agents read, write and run cells while your team edits live alongside them. You connect databases via SQLAlchemy, and compute comes from local Docker or from further machines over SSH. Teams, roles, audit logging and a command-line client are included.
Who it suits
Data teams with technical skills who want to do analysis together with agents without handing their data to a third-party platform. If you don’t want to run anything yourself, the hosted Alkera platform is the alternative.
Data protection for businesses
When self-hosted, files and notebooks stay on your servers. Agents use your own keys from Anthropic, OpenAI or Amazon Bedrock, so inputs go to those providers and fall under their terms. For the hosted platform, the privacy policy allows storage anywhere in the world. EU data residency and zero data retention are available only in the Enterprise plan. The vendor does not explicitly mention a data processing agreement for customers, nor whether inputs are used for training.
Limits
Installation and operation require Docker and server skills. The project is very young and shows little usage on GitHub so far. Without your own API key, no agents run.