Use Rowset as a stable backend for user-owned structured datasets. Rowset can be
used through MCP, the Rowset CLI, or the REST API. Treat these as peer
interfaces over the same account and data. During setup, prefer native remote
MCP only with private bearer-secret configuration, fall back to the CLI for
trusted terminal or local-file workflows, and use REST for code-only or
HTTP-only runtimes.
This skill is the durable guide for working with Rowset after access has been
configured. Use
when the agent still needs to inspect its runtime,
select and configure an interface, store credentials, verify authentication, or
complete first-run activation.
Rowset evolves quickly, so do not treat static examples as a complete feature
or command reference. Use live capabilities and interface documentation for
current schemas, commands, endpoints, and workflows.
Discovery is task-driven, not a startup checklist. Do not load capabilities or
list datasets merely because a session started. Use the current live resources
only when the task needs them rather than maintaining a static feature catalog:
When the user supplies a dataset key or URL, inspect that dataset directly. MCP
accepts either value. For CLI or REST, extract the dataset key from
the URL before using
or
/api/datasets/{dataset_key}
. If the
relevant dataset is unknown, search with an explicit limit of 3, select one, then
inspect its current schema, instructions, metadata, relationships, and index
semantics. Search before creating duplicates. If no stable business key exists
for a new dataset, use Rowset's current generated-ID path documented by the live
capability guide.