Skip to Content
Agent ActionsFunction actions
On this page

Function Agent Actions

Publish an owned Function. Site Member Functions must be linked to this app; unlinked public and MiniUp User Functions are also available. In Agent Actions → Functions → Configure, set its name, title, description, risk, HTTP method, path, and input schema, then enable exposure.

For a generate_customer_summary Function, an input schema might be:

{ "type": "object", "properties": { "customerId": { "type": "string", "minLength": 1, "maxLength": 128 } }, "required": ["customerId"], "additionalProperties": false }

Inputs are sent as JSON for POST, PUT, PATCH, and DELETE; GET inputs become query parameters. Configure a fixed Function path, such as /summary. Agents cannot choose another Function or route through that input.

Schemas support object properties, arrays and items, primitive types, nullable types, required fields, enums, constants, descriptions, and numeric/string/array/object size limits. Objects must set additionalProperties: false. External references, recursive schemas, regular expressions, and composition keywords are not supported. Schemas are limited to 16 KiB and eight nested levels.

Functions default to Significant consequences. Select Read only only when the configured operation cannot change state. Ordinary reversible changes use Can change data; deletion uses Destructive. Descriptions and risk labels help agents understand the action but cannot enforce business rules inside author code.

Calls use the existing Function endpoint and identity proxies. Site Member access preserves MINIUP_USER, member roles, MINIUP_CALLER, quotas, and caller-scoped Function Memory. MiniUp User access retains its existing account caller identity. Public Functions stay public. API-key-only Functions need an eligible identity access mode before exposure; MiniUp never embeds a Function key in the browser.

MiniUp · Agent Actions · functions