Pipefy Database Tables

Database Tables

Tables, records (rows), schema columns (table fields), and attachments for Pipefy Database Tables. 17 MCP tools.

Cross-cutting patterns

• Same conventions as pipe building: introspect_type on inputs such as CreateTableFieldInput / UpdateTableFieldInput, debug=true on mutations.
• Pagination: get_table_records and find_records support first / after. With the unified MCP envelope, read top-level pagination.has_more and pagination.end_cursor (and pagination.page_size) and pass after=end_cursor for the next page (default page size is 50).
• find_records over paginated get_table_records when you know the field value. One find_records call with a column_id/search_value filter beats N pages of get_table_records.
• Legacy mutation envelope. Several table mutation tools still return the GraphQL operation name as a nested key under result (for example result.createTableRecord). Read the payload inside that key; shape may differ from tools that already use the unified envelope.

Table operations

Table field (schema column) operations

Record operations

Attachment uploads

Steps — find and update a record

1. Get table ID (if not known):

   MCP: get_tables organization_id=123

   CLI: pipefy table list

2. Find the record (use find_records, not pagination):

   MCP: find_records table_id=456 filter='{"column_id":"email","search_value":"user@example.com"}'

   CLI: pipefy record find --table 456 --filter '{"column_id":"email","search_value":"user@example.com"}'

3. Update one field (targeted):

   MCP: set_table_record_field_value record_id=789 field_id="status" value="Active"

   CLI: pipefy record update 789 --field-id status --value '"Active"'

   Update multiple fields:

   MCP: update_table_record record_id=789 node_fields='[{"field_id":"status","field_value":"Active"}]'

   CLI: pipefy record update 789 --fields '{"status":"Active"}'

Two-step destructive previews

Always call without confirm=true first, surface the preview (including confirmation_token) to the user, then call again with confirm=true and that token after explicit approval. CLI uses --yes (no token). Preview content per tool:

• delete_table — show table name, field count, and record count. Deleting a table destroys all rows and schema.
• delete_table_record — show record title and key field values so the user can identify which row will vanish.
• delete_table_field — show field name and type; warn explicitly that all column data will be permanently lost.

Never delete in a single call.

Success criteria

• get_table_records returns the created/updated records with correct field values.
• Schema changes reflect immediately in get_table.

Failure modes

• get_table_record / get_table_records omit empty fields. Records only return populated fields, so you cannot tell "field unset" from "field doesn't exist" without calling get_table for the full schema.
• create_table_record title silently overridden. When the first table field is a start-form-style label column, Pipefy uses that field's value as the record title, ignoring the title parameter. Don't rely on title if the first field auto-populates a label-like column.
• create_table_field rejects type: call introspect_type type_name="CreateTableFieldInput" for valid field types.
• find_records returns empty: check that column_id matches a field's ID (not label) from get_table.
• Pagination cursor expired: re-fetch from the beginning; cursors are short-lived.

See also

• skills/relations/pipefy-relations/SKILL.md — connect tables to pipes (and the table-relation ID namespace gotcha).
• skills/introspection/pipefy-introspection/SKILL.md — discover field input schemas.