describe_method
Returns the full I/O contract for a single access method -- always EXTENDED, meaning inputs, request body schema, and full output shape. This is where an agent gets column names: each row carries a param_type of input_required (a mandatory, exact-match WHERE predicate), input_optional, or output (a field a SELECT or RETURNING can reference). A resource has no single field list, because each method returns its own shape, so call this for the method chosen from list_methods before the first query against it.
See also:
[ MCP overview ] [ list_methods ] [ run_select_query ]
Inputs
| Argument | Required | Description |
|---|---|---|
provider | Yes | Provider name. |
service | Yes | Service under that provider. |
resource | Yes | Resource under provider.service. |
method | Yes | Method name as returned by list_methods. |
Output
| Renderer | Shape |
|---|---|
| KV | One row per input and output field: name, type, param_type (input_required, input_optional or output), shape and description |
Gating
Allowed in every server mode. Read only.
Example
Try this prompt with any elicitation-capable MCP client with the StackQL MCP server registered:
Show me everything I need to know to call the insert method
on Google Compute networks -- required parameters,
optional parameters, and the request body shape.
Here's an example in Claude Desktop: