Loading ToolBench
An MCP server for your Strapi CMS that provides access to content types and entries through the MCP protocol
Strapi MCP has 11 tools with visible definitions in src/index.ts. All tools have verb_noun naming (list_, get_, create_, update_, delete_, publish_, unpublish_, upload_). Descriptions are present and non-trivial (ranging 50-150 chars), meeting baseline minimums. However, critical gaps emerge: (1) Parameter descriptions are sparse or missing for complex objects like 'data' and 'filters', LLMs cannot infer what fields to pass; (2) No output schemas are documented, agents cannot know what fields to expect from responses; (3) Error handling provides no recovery guidance, errors are returned but lack actionable next steps; (4) Security concerns: file paths in upload_media_from_path expose path traversal risk without documented validation. (5) The 'filters' parameter on list_entries references 'Strapi filter syntax' but does not document that syntax, forcing agents to guess at format. Naming is consistent and action-oriented (good), but schema and error documentation severely lag production standards.
Create a new entry in a Strapi content type
Delete an entry from a Strapi content type
Get detailed information about a specific content type in Strapi
Get a specific entry from a content type in Strapi
List all content types available in Strapi
List entries of a specific content type in Strapi
Publish a draft entry in Strapi
No output schemas documented for any tool. Tools return data but agents cannot know field structure, types, or what to extract for downstream calls. Violates pattern:tool and pattern:response-shaper.
Parameter 'data' in create_entry and update_entry has type 'object' with no schema or field documentation. LLMs cannot determine what fields or structure to pass. Violates pattern:tool-description and pattern:constrained-input.
Parameter 'filters' on list_entries references 'Strapi filter syntax' but does not define what that syntax is. LLMs must guess the format (JSON? query string? special operators?). Violates pattern:tool-description.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 27 | - | v1 |
Unpublish a published entry in Strapi
Update an existing entry in a Strapi content type
Upload a media file to Strapi (base64 encoded, max 1MB)
Upload a media file to Strapi from a file path (max 10MB)
No error handling guidance. Tools return errors but do not tell LLMs what to do next (retry? ask user? use alternative tool?). Violates pattern:recovery-guide and pattern:error-classification.
upload_media_from_path accepts a file path parameter with no validation. Risk of path traversal attacks if LLM constructs malicious paths. Violates pattern:tool-gateway and pattern:secret-injection.
No pagination guidance in tool descriptions. list_entries accepts 'limit' and 'offset' but does not state max limit or advise agents on pagination strategy. Violates pattern:paginated-result.
No distinction between required and optional parameters clearly marked in tool descriptions. LLMs may omit optional params expecting defaults, but if defaults differ from expectations, silent failures occur. Violates pattern:tool-description.
delete_entry has no confirmation mechanism. Agents can invoke permanent deletion without a dry-run or approval step. Violates pattern:confirmation-request.
Parameter descriptions are generic or incomplete. E.g., 'The data for the new entry' does not explain what fields are required, allowed, or their types. Violates pattern:tool-description.