MCP server that provides access to Webflow's Data API and Designer API, enabling tool-driven interactions with Webflow sites, collections, pages, components, and more.
The server exposes 35 tools with basic structure but significant quality gaps. Tool names follow verb_noun patterns (e.g., sites_list, cms_collection_item_create) which is positive. However, ~60% of tools have incomplete or missing parameter descriptions. Input schemas are present for many tools but lack depth, most show only siteId and collectionId with minimal context. Several tools (e.g., cms_collection_item_create, de_element_create, de_element_update, webhooks_create) accept 'fieldData' or similar object parameters with type:object and no nested schema definition, making it impossible for LLMs to know what fields are valid. Descriptions are present but often generic (e.g., 'Create a new item in a collection', 17 characters, below the 10-1024 target but borderline). No visible error handling guidance, no output schema documentation, and no indication of idempotency or state-modification semantics. Tool composition is reasonable (single responsibility, chainable via siteId), but the lack of detailed parameter specs and output documentation severely limits LLM usability.
Query Webflow AI for assistance with Webflow API-related questions
Create a new item in a collection
Delete an item from a collection
Update an item in a collection
List items in a collection
List all collections in a site
List all comments for a page or resource
Object parameters lack nested schema definitions. Tools like cms_collection_item_create, de_element_create, de_element_update, de_style_update, de_asset_upload, de_variable_create accept 'fieldData' or similar objects typed only as 'object' with no nested properties documented. LLMs cannot determine valid fields, leading to invalid calls and wasted retries.
No output schemas documented for any tool. The rubric requires structured output documentation so LLMs know what fields to expect and can plan downstream tool calls. This is a complete gap across all 35 tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 49 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 54 | - | v1 |
Get details of a specific component
List all components in a site
List all assets in a Designer site
Upload an asset to a Designer site
Get details of a specific Designer component
List all components in a Designer site
Create a new element in the Designer
Delete an element in the Designer
Update an element in the Designer
Get details of a specific Designer page
List all pages in a Designer site
Update styles for an element in the Designer
Create a new variable in the Designer
Delete a variable in the Designer
List all variables in a Designer site
Update a variable in the Designer
List all domains for an enterprise site
Share URL for connecting to Webflow Designer App locally. Only valid for OSS MCP Version.
Get details of a specific page
List all pages in a site
Update a page's meta information
Update a page's properties
List all scripts registered for a site
Register an inline script for a site
List all sites accessible to the authenticated user
Create a webhook for a site
List all webhooks for a site
List all workflows for a site
Destructive tools (delete operations: cms_collection_item_delete, de_element_delete, de_variable_delete) lack confirmation or dry-run semantics. Per the pattern, irreversible operations should support a confirm_before_execute step or clear warnings in the description.
No error handling guidance visible. Tools do not return recovery suggestions, error classifications, or actionable feedback. E.g., if a siteId is invalid, what should the LLM do? No guidance provided.
ask_webflow_ai has no input schema visible. Without a schema, the LLM cannot determine what parameters to pass. This tool scores 0 on schema dimension.
List tools (sites_list, cms_collections_list, etc.) lack explicit pagination parameters (limit, offset, page_size) and no mention of result caps in descriptions. If a collection has 1000 items, does sites_list return all or paginate? Unclear.
Descriptions are generic and brief. Examples: 'List all sites accessible to the authenticated user' (57 chars), 'List all collections in a site' (30 chars), 'Create a new item in a collection' (33 chars). These lack context on WHEN to use each tool, dependencies, or prerequisites. Per the rubric baseline (194 chars avg for A+ tools), these are significantly underdeveloped.