The read path runs spaces, then a node search, then records, with pagination, field filtering and sorting available on the way down. Writing is the part worth having: the schema of the target datasheet goes to the model first, so a new record fills a known structure rather than guessing at column names. Attachments arrive by web URL rather than from local disk, and the storage information that comes back is what gets attached to the row.
A server over AITable workspaces — spaces, datasheets, forms, dashboards and folders — that reads records out of a datasheet and creates new ones in it.
- Every workspace the authenticated user can open, plus a node search across datasheets, forms, dashboards and folders by type and query.
- Paginated record reads with field selection, formula filtering and sort order.
- The JSON schema of a datasheet's fields, giving the model the structure a row is expected to have before it writes one.
- Record creation that extracts values from user-provided text and maps them onto that schema.
- Attachment upload from a web URL, returning storage information that a record can be created with.
An AITable account with permission on the spaces you want to reach; every tool acts as the currently authenticated user.
One command plus a key — npx -y @apitable/aitable-mcp-server, then supply credentials