Percepta MCP Server
An AI-driven platform for frontend semantic cognition and automation.
Percepta MCP Server
基于模型上下文协议(MCP)的强大工具集合,提供浏览器自动化、AI分析、视觉处理、网页爬虫、自动化测试生成和DevTools分析等功能。
本项目专为本地单用户使用设计,不适合公开部署。
快速开始
环境要求
- Python 3.12+
- uv (推荐) 或 pip
本地安装运行
# 克隆项目
git clone <repository-url>
cd percepta-mcp
# 安装依赖 (推荐使用 uv)
uv venv
uv sync
# 配置环境变量
cp .env.example .env
# 编辑 .env 文件,添加您的 API 密钥
# 安装浏览器驱动
uv run playwright install chromium
# 启动服务
uv run python -m src.percepta_mcp.server
Docker 部署
# 配置环境变量
cp .env.example .env
# 编辑 .env 文件
# 构建并启动
docker-compose up -d
# 查看状态
docker-compose ps
curl http://localhost:8000/health
MCP 客户端配置
Claude Desktop
编辑配置文件 claude_desktop_config.json:
本地运行配置:
{
"mcpServers": {
"percepta-mcp": {
"command": "uv",
"args": ["run", "python", "-m", "src.percepta_mcp.server"],
"cwd": "/path/to/percepta-mcp"
}
}
}
Docker 配置:
{
"mcpServers": {
"percepta-mcp": {
"command": "docker",
"args": ["exec", "percepta-mcp", "uv", "run", "python", "-m", "src.percepta_mcp.server"]
}
}
}
VS Code
配置 settings.json:
{
"mcp.servers": {
"percepta-mcp": {
"command": "uv",
"args": ["run", "python", "-m", "src.percepta_mcp.server"],
"cwd": "${workspaceFolder}"
}
}
}
功能特性
- 浏览器自动化: 页面导航、元素交互、截图、表单操作
- AI 智能分析: 支持 OpenAI、Anthropic、Google、Ollama
- 视觉处理: 图像分析、OCR、对象检测
- 网页爬虫: 智能数据提取、结构化爬取
- 自动化测试: AI 驱动的测试用例生成和执行
- DevTools 分析: 性能监控、异常检测
项目结构
percepta-mcp/
├── src/percepta_mcp/
│ ├── server.py # MCP 服务器主入口
│ ├── config.py # 配置管理
│ ├── ai_router.py # AI 提供商路由
│ └── tools/ # 工具模块
├── tests/ # 测试文件
├── .env.example # 环境变量模板
├── docker-compose.yml # Docker 部署配置
└── Dockerfile # Docker 镜像构建
测试
# 运行所有测试
uv run pytest
# 带覆盖率报告
uv run pytest --cov=src --cov-report=html
当前测试覆盖率: 89% (209个测试全部通过)
环境变量
关键环境变量配置:
OPENAI_API_KEY: OpenAI API 密钥ANTHROPIC_API_KEY: Anthropic API 密钥GOOGLE_API_KEY: Google AI API 密钥PERCEPTA_HOST: 服务监听地址 (默认: 0.0.0.0)PERCEPTA_PORT: 服务端口 (默认: 8000)PERCEPTA_LOG_LEVEL: 日志级别 (默认: INFO)
许可证
MIT License
Verwandte Server
Alpha Vantage MCP Server
SponsorAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Remote MCP Server (Authless)
A remote MCP server deployable on Cloudflare Workers without authentication.
NestJS MCP Server Module
A NestJS module for building MCP servers to expose tools and resources for AI, with support for multiple transport types.
CAD-MCP
Control CAD software with natural language instructions to perform drawing operations.
FrankenClaw
Modular MCP toolbox that gives AI agents controlled access to shell, files, Git, Ollama, Shopify, and more — without losing cost or model control.
AgentSeal
Every agent action is recorded in a SHA-256 hash chain; with this, you can actually prove to clients that your agent did what it said it did
AppControl Windows System History
An MCP server that gives AI agents like Claude, Cursor, and Gemini access to historical Windows CPU, GPU, temperature, and privacy data gathered by the AppControl Windows app.
Berry MCP Server
A universal framework for easily creating and deploying Model Context Protocol servers with any tools.
REI Crypto MCP Server - Beta
A FastMCP implementation of the in-house MCP servers used by Units of the Rei Network.
widemem.ai
Open-source AI memory layer with importance scoring, temporal decay, hierarchical memory, and YMYL prioritization
tachibot-mcp
Stop AI Hallucinations Before They Start Run models from OpenAI, Google, Anthropic, xAI, Perplexity, and OpenRouter in parallel. They check each other's work, debate solutions, and catch errors before you see them.