# @validpilot/ai-verify-mcp [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/validpilot/ai-verify-mcp  
**GitHub Stars:** 1  
**npm Downloads (last month):** 274  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/validpilot-ai-verify-mcp

## Description
AI 编程验证平台。让 AI 代码生成结果可验证、可信赖。MCP 协议原生支持，证据驱动。

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "ai-verify-mcp": {
    "command": "npx",
    "args": ["-y","@validpilot/ai-verify-mcp"]
  }
}
```

## Documentation

## What @validpilot/ai-verify-mcp does

@validpilot/ai-verify-mcp provides MCP tools for checking AI-generated web applications and recording the results. Its browser capabilities cover opening pages, interacting with controls, locating elements, inspecting state, capturing screenshots, and checking performance. Validation features include URL, text, and element assertions; accessibility checks using axe-core; security scans; network monitoring; and analysis of browser errors.

The server is designed around evidence-backed verification. Actions can produce screenshots, console records, network-request data, and other artifacts. These results can be assembled into structured JSON and Markdown reports, with diagnostic output that includes likely error causes, confidence information, and suggested fixes.

## How it works

An MCP-compatible AI client calls the server’s tools as it works through a validation task. A typical flow opens a page, fills fields, clicks a control, checks the resulting state, and asserts that the expected URL or content appears. Evidence is collected throughout the flow so a later reviewer can inspect what happened at each stage.

The package also includes MCP Prompts for common workflows:

- `/validate-login`
- `/submit-form`
- `/audit-performance`
- `/audit-security`
- `/visual-regression`
- `/debug-page`
- `/e2e-flow`

These prompts provide an entry point for orchestrating multiple low-level tools. Additional skill guidance is distributed with the npm package, while a Trae-specific skill extension is available as an optional enhancement.

## Setup and configuration

Install the package with npm and start the MCP server with:

```bash
npm install @validpilot/ai-verify-mcp
npx @validpilot/ai-verify-mcp start
```

It can also be used directly from the command line for a website validation, a named screenshot, or a quick check:

```bash
npx @validpilot/ai-verify-mcp validate --url https://example.com
npx @validpilot/ai-verify-mcp screenshot --url https://example.com --name evidence-001
npx @validpilot/ai-verify-mcp quick-check --url https://example.com
```

MCP configurations shown for Cursor and Claude Code launch the package with `npx`. HTTP mode can be selected with `MCP_MODE` and a listening port with `MCP_HTTP_PORT`; Windsurf can use the basic `npx` configuration. After changing an installed version or MCP configuration, restart or reload the client.

## Tools and capabilities

The toolset includes capabilities for:

- Browser navigation, interaction, element location, state inspection, debugging, overlays, CAPTCHA handling, and sessions.
- Screenshots, visual checks, evidence collection, chain specifications, and validation reports.
- Assertions for elements, text, and URLs, plus smoke tests and counterfactual analysis.
- Accessibility checks, performance analysis, network and asset inspection, and CSS variable tracking.
- Security header checks, OWASP Top 10 checks, security scanning, and API probing.
- Error analysis, trace correlation, MCP diagnostics, contracts, and skill validation.

Tool aliases preserve compatibility with older tool names. The server’s workflow prompts and reporting features are useful when the agent needs both execution and an auditable record of the result.

_Full upstream README: https://allmcps.com/mcp/validpilot-ai-verify-mcp/readme_