by LakehouseBox SLCommunitySave tables AgentOne can query again later with SQL, and add rows over time. Apache Iceberg, run in Europe.
LakehouseBox lets AgentOne keep data between chats. AgentOne can save results as tables, add rows over time, and query them again later with SQL.
LakehouseBox is an Apache Iceberg lakehouse run in Germany. AgentOne can see every catalog and table you have access to, with row counts, size and when each was last updated. It can search tables by name, column or documentation, run DuckDB SQL, including joins across catalogs, and, on catalogs where you grant write access, create tables and add, change or remove rows. The write tool is marked as able to delete data, so AgentOne asks before running it. Queries run on LakehouseBox's shared engine when the agent's environment can't reach LakehouseBox (most chat sandboxes). Agents that can reach LakehouseBox compute in their own DuckDB. AgentOne can also read files stored next to your tables, such as images, and open public catalogs that other organizations publish.
If you have no account yet, AgentOne can set one up for you when you connect. The account includes one catalog, up to 1 GB. Add your email within 72 hours to keep it. Otherwise it is deleted with its data.
Catalog owners can add notes (AGENTS.md) and column docs that explain what the data means and how tables join. Every tool runs with the access you approve when connecting. Your data stays in open Iceberg tables that you can also query from DuckDB, Spark or Snowflake.