Loading ToolBench
A Next.js application with an integrated WordPress MCP (Model Context Protocol) client for managing WordPress sites through an AI chatbot interface. Provides tools for WordPress content management, media handling, user management, and WooCommerce operations.
This is a TypeScript-based AI chatbot client that BRIDGES to WordPress via MCP, not an MCP server itself. The repository shows a Next.js application with AI SDK integration and Firecrawl-based web scraping tools. Critical issues: (1) Tools 1-5 (analyzeWebContent, analyzeCompetitor, searchContentIdeas, extractStructuredData, batchAnalyzeUrls) are AI SDK tool definitions, NOT MCP-compliant tool definitions, they lack MCP registration, protocol metadata, and proper schema documentation. (2) Tool 6 (WordPress dynamic tools) is inferred but never explicitly registered in visible code; the actual MCP server stub at tools/wp-connector/package.json is incomplete ('TODO: implement MCP server'). (3) No input schemas visible for any tool, only partial parameter hints in the tool descriptions. (4) Descriptions are present but generic; parameter descriptions are sparse. (5) The codebase uses deprecated AI SDK patterns (ai@5.0.0-beta.6) rather than production MCP protocol. This is NOT a production MCP server, it's a client application that will eventually connect to a WordPress MCP server, but that server is not yet implemented.
Dynamically loaded tools from WordPress MCP server based on connection configuration. Tools are converted from MCP protocol to AI SDK format.
Analyze a competitor website to identify content gaps, topics, and strategic opportunities
Scrape and analyze a website for content strategy insights including SEO metrics, readability, and content structure
Analyze multiple URLs at once for content comparison and competitive analysis
Extract specific structured data from a webpage using AI-powered extraction
Search the web for content ideas and trending topics in a specific niche or industry
No MCP server implementation, tools/wp-connector package.json contains only a stub with 'TODO: implement MCP server'. The WordPress MCP server does not exist, only a client application.
Tools 1-5 are AI SDK tool definitions (firecrawl-tools.ts), not MCP-protocol definitions. No JSON Schema input validation visible; only parameter hints in descriptions. Missing required schema properties: type, properties, required fields.
Parameter descriptions are absent or minimal. Example: 'url' param in analyzeWebContent has no description; 'comparisonType' in batchAnalyzeUrls is typed as 'enum' but no enum values listed. LLMs cannot infer parameter constraints without explicit descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 28 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 39 | - | v1 |
'WordPress dynamic tools' tool definition is inferred, not explicitly registered. No schema, no parameter list, no fixed interface, tool capabilities depend on runtime MCP server connection that does not yet exist. Impossible to validate or test.
No error handling patterns documented. Tools like analyzeWebContent (web scraping) will fail silently or with generic network errors, no recovery guidance for LLM ('If URL is unreachable, try a different domain'). Per pattern:recovery-guide, errors must guide the agent.
Tool 'batchAnalyzeUrls' lacks documentation on how comparison results are structured. Is it a ranked list? A diff? A table? LLM cannot plan downstream actions without knowing the response shape.
No pagination or result limits documented for searchContentIdeas (defaults to limit=5, but what if an LLM changes it?). Per pattern:paginated-result, tools returning lists must accept page/offset and limit and document max result size.