Tools reference
Private beta: access through the IAMP platform only
The Quable MCP server is a private beta and not a public API. Access is only authorized through the IAMP platform: any other access is not authorized. Any tool, prompt, resource, parameter or behavior can change or be removed at any time, without notice.
At the time of writing, the Quable MCP server exposes 35 tools, 2 prompts and a few resources. Your AI assistant reads the description of each tool and chooses which one to call: you normally don't call tools yourself, but knowing them helps you write precise requests and understand what the assistant does.
Conventions
- Codes, not labels. Every identifier parameter expects the exact code (
fr_FR,reference_color,completeness_e_commerce), never a display label. Thelist_*andsearch_*tools resolve labels to codes. localeon most tools sets theX-Quable-Localeheader: the language of translated labels in the response and the locale in which filters are evaluated. It does not select which data languages are written or returned.- JSON parameters. Parameters described as JSON string expect a serialized JSON object or array, e.g.
'{"fr_FR": "Rouge"}'. - marks a write tool. Each one tells the assistant to confirm with you before calling it, and fails with a Read access token.
Overview
Session and server
test_credentials
Checks that the instance name and token sent in the headers are valid. Use it to test a new connection.
Parameters: none.
Returns: {"status": "valid credential"}, or an authentication error.
get_user_info
Returns the profile of the token's user. Call it at the start of a session so the assistant knows which locale to use by default.
Parameters: none.
Returns: localesAvailable (active data locales), localeDefault, localeI18nDefault (interface language), permissions (permissionFeatures, permissionField, permissionCatalog, damPermissionCatalog), admin, active, timezone, username, email, firstName, lastName, dateLogged.
server_time
Returns the current UTC time of the server. Does not call your instance and needs no authentication.
Returns: {"utc_now": "2026-11-19T10:42:00+00:00"}
get_pim_url
Builds a clickable link to a product or asset page in the Quable PIM interface.
Returns: {"url": "https://your-instance.quable.com/#classification/product-new/12345/edit"}. A missing or non-numeric legacy_id returns an error: the tool never guesses a link.
skill_word_count
Counts the visible words of a text. HTML tags, comments and scripts are removed and entities decoded first, so a rich-text value counts only what a reader sees. Useful to check the length of a generated description. Needs no authentication.
Returns: {"word_count": 128}
Catalog structure
These tools tell the assistant how your PIM is configured. Most of them are called by the pim-bootstrap
list_locales
Lists the active data locales of the instance.
Parameters: none.
Returns: a list of {id, label, fallback?}. id is the code to use in other tools (e.g. fr_FR). fallback is the locale used when a value is missing.
list_document_types
Lists document types (product types) and link types.
Returns: {document_types: [{id, name, object_type, has_variant}], link_types: [{id, name}]}. has_variant tells whether the type accepts variants. Link type ids are used by add_link and remove_link.
list_completenesses
Lists the completeness rules configured on the instance. It does not return scores: read them on a product with get_product (view completeness) or filter with search_products.
Returns: a list of {id, name, document_type, required_attributes, required_links: [{id, minimum}]}.
list_attributes
Finds attribute codes, for example from an approximate name.
Returns: a list of {id, name, type, documentTypeId, group}. Types simple_select and multi_select are select attributes (also called combo and multicombo): use list_predefined-values
list_attributes works on document types only, not on link types.
search_classification
Resolves classification (category) ids from a name or id.
Returns: {items: [{id, label, path, parent_id, catalog_id, active, legacy_id}], total, returned, offset, missing_ids}. Labels are not unique: the same name can exist under several parents, so check path.
search_attribute-sets
Lists attribute sets, i.e. which attributes apply to the products of a document type and count for completeness.
Returns: {items: [{id, code, name, document_type_id, attributes}], total, returned}. attributes lists the attribute codes of the set.
update_attribute-set
Creates or updates an attribute set. Idempotent by code.
Data model change
An attribute set controls which attributes are visible on every product of a type, cascades to variants and changes completeness scores. Inspect the current state with search_attribute-sets first.
Products
get_product
Reads one product (document) by its exact id. To read several known products, the assistant calls it once per id rather than searching.
Returns: the basic view (id, documentType, active, attributeSet, dateCreated, dateModified, mainAssetThumbnailUrl, legacy_id) plus the requested sections, and a _meta block:
{
"id": "REF-001",
"attributes": { "product_name": { "fr_FR": "Veste", "en_GB": "Jacket" } },
"completenesses": { "completeness_e_commerce": { "fr_FR": 85, "en_GB": 100 } },
"_meta": {
"views": { "used": ["basic", "attributes", "completeness"], "omitted": ["links", "assets", "..."] },
"output_locales": { "used": ["fr_FR", "en_GB"], "omitted": [] }
}
}
Section keys in the result: completeness → completenesses, links → documentLinks, assets → assetLinks, workflow → workflows. In _meta, omitted means not requested, not does not exist.
search_products
Searches and filters products when their ids are not known.
Set at most one of sort_by, sort_by_completeness and sort_by_attribute.
Returns: {items, total, limit, offset, _meta}.
Example: the 10 products with the best Spanish e-commerce score, without filtering:
{
"sort_by_completeness": "completeness_e_commerce",
"order_direction": "desc",
"locale": "es_ES",
"limit": 10
}
Attribute filters
Used by search_products, search_asset and search_variants:
create_product
Creates a product with its classification, attribute set and values.
Invalid input is refused before anything is created.
Returns: {status, message, document, missing_required, not_persisted}. missing_required lists required attributes still empty. not_persisted lists values the PIM did not store: retry them with update_product.
update_product
Partially updates a product: only the keys you send change. If the PIM does not persist a value (for example a character limit is exceeded), the tool returns an error instead of reporting success.
Top-level keys of the payload:
The document type cannot be changed after creation.
{
"attributes": {
"product_description": { "fr_FR": "Nouvelle description", "en_GB": "New description" },
"product_color": [{ "id": "red" }],
"product_weight": 1.2
}
}
Returns: {status, message, document}, where document has the same shape as get_product.
Attribute value formats
To clear a value, send null or an empty string.
Variants
get_variant
Reads one variant (SKU) with every attribute of the variant type.
Returns: {found, variant_code, legacy_id, product_code, name, active, sequence, date_created, date_modified, attributes, missing_required, assets: [{asset_id, link_type, sequence}]}. attributes lists every attribute allowed by the parent's attribute set, null when never set. An unknown code returns found: false instead of an error.
search_variants
Lists the variants of a product, or searches variants across the catalog.
Returns: {items: [{variant_code, legacy_id, product_code, name, active, sequence, attributes}], total, limit, offset}.
create_variant
Creates a variant under an existing product. The product's type must accept variants (has_variant in list_document_types).
A variant follows its parent's attribute set: values of attributes outside that set are refused.
Returns: {status, message, variant: {id, legacy_id, product_code, name, sequence, active}, missing_required, not_persisted}.
update_variant
Changes attribute values of a variant. The code, parent and status are never changed.
- A select is replaced by the full list sent, never merged.
[]ornullclears it. - A translatable text only changes the locales sent.
- Values the PIM would drop or garble are refused before writing, and the result is read back.
Returns: {status: "updated" | "partially_updated", message, variant_code, product_code, attributes, not_persisted, missing_required}.
Assets
get_asset
Reads an asset's metadata: name, original filename, size, thumbnail and preview URLs, attributes, classifications and links. No file content.
get_asset_file
Downloads an asset's content so the assistant can read it.
What you get, by file type:
Other file types are rejected: use get_asset for their metadata and URLs. See size limits.
search_asset
Searches DAM assets.
Returns: a compact paginated list. Each item has a legacy_id for get_pim_url.
update_asset
Partially updates an asset's attributes.
For select attributes on assets, prefer the shape returned by get_asset: an array of {id, labels}, where id is the option code.
{
"qua_asset_non_translatable_text": "Label",
"qua_asset_textfield": { "fr_FR": "texte", "en_GB": null },
"qua_asset_multicombo": [{ "id": "season_winter", "labels": { "en_GB": "Winter", "fr_FR": "Hiver" } }]
}
Returns: {status, message, asset}.
Select options
list_predefined-values
Lists the options of a select attribute (types simple_select / multi_select). Use it to find the code behind a label before filtering or updating, or to check that a code is free.
Returns: a list of {code, attribute_id, labels, label, description, legacy_id}.
update_predefined-value
Creates or updates an option of a select attribute. Idempotent: creates the option when the code is new, updates its labels otherwise.
Returns: {status, message, predefined_value}, where status says whether the option was created or updated.
Links
add_asset_to_product
Links an asset to a product. The product is the link origin, the asset the target.
remove_asset_from_product
Removes the link between an asset and a product.
add_link
Creates a link of any type between two objects: product → product (cross-sell, upsell, replacement), product → asset, etc. One link per call.
If the link already exists, returns already_linked: true without creating a duplicate.
remove_link
Removes a link of any type. One link per call. Give either link_code, or origin + target + link_type.
If the link no longer exists, returns already_removed: true without an error.
Exports
Exports are asynchronous and follow four steps. See the export workflow.
list_export_profiles
Lists the export profiles saved in the PIM, with their UUID for trigger_export.
get_export_options
Step 1: discovers what can be exported.
Returns: the allowed export models, plus columns and format options when requested.
trigger_export
Step 2: starts an export and returns a process_id immediately. Give exactly one of export_profile_id or export_profile.
Inline profile, all columns (default):
{ "model": "document.qua_document", "extension": "csv", "locales": ["en_GB", "fr_FR"] }
Inline profile with a column selection (only when you ask for specific columns or headers):
{
"model": "document.qua_document",
"extension": "csv",
"locales": ["fr_FR"],
"mappings": [
{ "attributeCode": "document_code", "attributeType": "unlocalized_text", "name": "document", "position": 0 }
]
}
For localized attributes in mappings, add the locale in parentheses after the code, e.g. "qua_textfield (fr_FR)". Use get_export_options with model to get the exact column codes and types.
get_export_status
Step 3: polls an export.
Returns: {process_id, status, files: [{id, status}], startDate, endDate, hint}. status is pending, running, finished, failed or canceled. hint tells the assistant what to do next: poll again in about 10 seconds, download the file, or report that the file was delivered to the remote location or that the export failed.
get_export_file
Step 4: downloads the exported file.
Quable delivers exports as a zip archive: the tool unzips it and returns the CSV or XLSX file itself (filename, mimetype, base64 content). Files above 250 KB return metadata only, with too_large: true. In that case, reduce the scope (fewer columns, locales or products), download the file from the Quable interface, or use upload_location.
Prompts
Prompts are ready-made instructions the agent can load.