Record reads and writes

This is direct access to the rows in your Xata tables. A Beam agent reads records to answer a step in your workflow, and when a rule you approved calls for it, writes new rows or edits existing ones through the connection your account allows. Every change is logged with what triggered it. The agent respects the field types and constraints your schema defines and will not push values that break them. Edits that would overwrite fields outside its rule, or that hit a constraint it cannot satisfy, are stopped and routed to a person to handle safely.

Full-text search queries

This finds records by searching across your Xata data rather than by key alone. When a workflow needs a match, the agent runs the search your rule defines, reads the results, and uses the top match to continue or to update a linked record. It records which query ran and what it returned. When a search comes back with several equally strong candidates, or nothing at all, the agent does not pick blindly; it holds the step and asks a person which record is right. This keeps automated lookups honest when the data is genuinely ambiguous.

Schema-aware updates

This holds writes in line with how your Xata tables are structured. Before an agent changes a record, it reads the schema, measures the update against the types and required fields, and applies the change only when it fits your approved rule. Valid edits are saved and logged. Changes that would violate a constraint, touch a column outside the rule, or need a schema alteration are refused and passed to a person. The agent will not migrate or restructure tables on its own, since changing the shape of a database is a human decision, never an automatic one.