S22 Web Agent MCP Server
Provides MCP tools for web page scanning, data extraction, local storage, markdown report generation, and browser inspection using Playwright on a Samsung S22.
README
S22 Web Agent
A mobile-first web automation and MCP portfolio project running on a Samsung S22 using Termux, Node.js, SQLite, Express, MCP, and Debian proot Playwright.
This project explores whether an Android phone can act as a lightweight AI automation development environment, local tool server, and browser automation node.
Project Goal
The goal of this project is to build a practical AI automation agent that runs from a Samsung S22 instead of a VPS or laptop.
It demonstrates how a phone can host:
- CLI automation
- HTTP API tools
- MCP tools
- SQLite-backed local memory
- Markdown report generation
- Browser inspection through Chromium and Playwright inside Debian proot
Why Samsung S22?
This project is intentionally built on Samsung S22 through Termux.
Instead of using a VPS or laptop as the main runtime, the phone acts as the local development and execution environment.
This makes the project useful as a portfolio demonstration for:
- AI automation engineering
- MCP tool development
- Applied AI workflow prototyping
- Mobile Linux experimentation
- Browser automation with constrained hardware
Current Features
- Scan a webpage URL from CLI
- Scan a webpage URL through HTTP API
- Normalize simple domain inputs such as
example.comintohttps://example.com/ - Extract page title, description, headings, and links
- Save scan result into SQLite
- List saved scan records
- Show full scan detail by ID
- Delete unwanted scan records
- Generate markdown report from saved scan data
- Run an MCP server
- Expose scan/list/page/report functions as MCP tools
- Inspect browser-rendered pages through MCP using a Debian proot Playwright worker
- Track development progress using Git
Tech Stack
- Samsung S22
- Termux
- Debian proot
- Node.js
- npm
- Cheerio
- Express
- SQLite
- MCP SDK
- playwright-core
- Chromium
- Git
Architecture
Current working flow:
ChatGPT / MCP Inspector
|
v
MCP Server on Samsung S22
|
v
HTTP API Server on port 3001
|
v
Scanner + SQLite database
Browser inspection flow:
MCP browser_inspect_url
|
v
Debian proot Playwright worker on port 3002
|
v
Chromium headless browser
The CLI and HTTP API share the same scanner module.
CLI scan -> src/scanner.js
API POST /scan -> src/scanner.js
The browser inspection tool uses a separate Playwright worker running inside Debian proot.
CLI Usage
Install dependencies:
npm install
Scan a webpage:
node src/index.js scan https://example.com
Simple domain input is also supported:
node src/index.js scan example.com
List saved scans:
node src/index.js list
Show scan detail:
node src/index.js show 1
Delete scan record:
node src/index.js delete 1
Generate markdown report:
node src/index.js report 1
Show help:
node src/index.js help
Example CLI Flow
node src/index.js scan example.com
node src/index.js list
node src/index.js show 1
node src/index.js report 1
HTTP API Server
The project includes an Express-based HTTP API server running inside Termux on the Samsung S22.
Start the server:
npm run server
Default API port:
3001
Available endpoints:
GET /health
GET /pages
GET /pages/:id
POST /scan
GET /report/:id
Example health check:
curl http://localhost:3001/health
Example scan request:
curl -X POST http://localhost:3001/scan \
-H "Content-Type: application/json" \
-d '{"url":"https://example.com"}'
Simple domain input is also supported:
curl -X POST http://localhost:3001/scan \
-H "Content-Type: application/json" \
-d '{"url":"example.com"}'
Example LAN access from PC:
http://<S22-IP>:3001/health
http://<S22-IP>:3001/pages
This confirms the S22 can act as a local network-accessible tool server.
MCP Tools
The project includes an MCP server that exposes the scanner and browser inspection features as tools.
Tested MCP tools:
job_radar_scanjob_radar_list_pagesjob_radar_get_pagejob_radar_get_reportbrowser_inspect_url
Start the MCP server:
npm run mcp
Example MCP Inspector usage:
npx @modelcontextprotocol/inspector npm run mcp
Debian Proot Playwright Worker
Some websites require browser-rendered inspection instead of static HTML scanning.
For that case, the project uses a Debian proot worker running Chromium through Playwright.
Default worker port:
3002
The proot Playwright setup is documented here:
docs/proot-playwright.md
Worker source files are stored here:
tools/proot-playwright-worker
URL Normalization
The scanner supports simple domain input.
Example:
example.com
is normalized to:
https://example.com/
This works through:
- CLI
- HTTP API
- MCP tool calls
Documentation
Additional project notes:
docs/progress.md
docs/portfolio-polish.md
docs/proot-playwright.md
Limitations
- Some websites may block simple HTTP requests with 403 Forbidden.
- The CLI and HTTP API scanner currently use static HTML extraction.
- JavaScript-heavy websites require the separate Debian proot Playwright worker.
- The browser worker must be started separately inside Debian proot.
- Termux or proot may slow down when the phone screen is off.
- The project is built for learning and portfolio demonstration, not production use.
Roadmap
Planned improvements:
- Add architecture diagram image
- Add screenshots for GitHub demo
- Add
.env.example - Add license file
- Add sample scan output
- Add screenshot capture metadata from Playwright
- Improve classifier scoring notes
- Improve error handling for failed scans
- Save browser inspection result to database
- Add portfolio demo report export
Portfolio Summary
Built a Samsung S22-hosted web automation and MCP prototype using Termux, Node.js, SQLite, Express, Debian proot, Chromium, and Playwright.
The system scans web pages, extracts structured page data, stores results locally, exposes results through CLI, HTTP API, and MCP tools, and generates markdown reports.
This demonstrates practical AI automation workflow development without relying on VPS infrastructure.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.