A WebMCP test agent that discovers tools from web pages, generates test scenarios using Gemini AI or rule-based inference, and executes automated tests against those tools. Includes an Electron-based UI for scenario generation and test execution.
This is a UI test automation tool, not an MCP server. It exposes 5 tools via Electron IPC, not via MCP protocol. Tool definitions lack rigor: 3 of 5 tools have no visible input schemas in the source code (document.modelContext.getTools, document.modelContext.executeTool, inferToolsFromDOM). Descriptions are present but generic (50-150 chars, below the 194-char baseline for production tools). Parameters lack type definitions and constraints. No error handling guidance. The tool set conflates concerns (discovery, execution, scenario generation) without clear composition. This server is not MCP-compliant and would not function as an MCP server.
Executes a registered WebMCP tool with the provided parameters and returns the result
Retrieves all registered WebMCP tools from the target page's document.modelContext, including their names, descriptions, and input schemas
Rule-based fallback scenario generator that creates test scenarios from JSON Schema without requiring external API. Generates happy path and violation scenarios based on schema constraints (pattern, enum, minimum, maxLength, minLength, required fields, type)
Generates test scenarios from a WebMCP tool schema using Google Gemini API. Produces 6-10 distinct test cases including valid (happy path) and invalid (boundary, format violation, missing required field) scenarios
Automatic tool inference from DOM elements. Reads HTML5 validation attributes (pattern, required, min, max, maxlength, type=email) to build JSON Schema and uses browser's form.checkValidity() for validation. Fallback for sites without native WebMCP
Not an MCP server. This is an Electron desktop app using IPC, not MCP protocol. Tools are exposed via ipcMain.handle(), not via MCP tool registration. Cannot be used as an MCP server by any MCP client.
Three tools (getTools, executeTool, inferToolsFromDOM) have no visible input schemas in source code.
Descriptions are generic and under 100 chars for 3 tools. Baseline is 194 chars. Descriptions lack WHEN to use, prerequisites, or recovery guidance. Example: 'Retrieves all registered WebMCP tools' (50 chars) does not explain when to call this vs other discovery tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
executeTool accepts 'ref' (object with name) and 'params' (object), but no type constraints, enums, or validation rules are documented. LLM cannot infer valid structure. Missing: required fields, nested schema for ref.name, param validation rules.
No error handling guidance. generateScenariosWithGemini returns {ok: false, reason: '...'} but does not categorize errors as retryable, user-fixable, or fatal. LLM cannot decide next action. Missing: recovery hints, actionable error messages.