mcpField
B
Bezlo MCP

Bezlo MCP

Local
github.com/gba3124Other0Updated today
Repository

About Bezlo MCP

Bezlo MCP is an agent layer for checking web pages in device-sized mobile frames. It is for teams testing mobile layouts and interactions, with diagnostics, multi-device runs, and evidence outputs through a server, CLI, and GitHub Action.

Key features

  1. 01Device-sized mobile screenshots
  2. 02Layout and page-error diagnostics
  3. 03Tap, type, and scroll actions
  4. 04Multi-device matrix runs
  5. 05JSON, JUnit, and video evidence

Use cases

  1. 01Check a page on iPhone and Android sizes
  2. 02Catch overflow and console errors
  3. 03Automate sign-in and other UI flows
  4. 04Attach screenshot and test reports to PRs

MCP support

Its stdio MCP server provides an agent-facing interface to Bezlo’s session and capture engine for inspecting and acting on web pages.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
  "mcpServers": {
    "gba3124-bezlo": {
      "command": "npx",
      "args": [
        "-y",
        "bezlo-mcp"
      ]
    }
  }
}
Run commandnpx -y bezlo-mcp

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmbezlo-mcp0.1.2

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Bezlo — open agent layer

Bezlo puts any web page inside a true-to-size iPhone / iPad / Android frame and lets an AI agent see it and act on it: screenshots, interactive-element lists, layout diagnostics (overflow, safe-area, touch identity), console/page errors, network mocks, tap/type/scroll flows, device matrix runs — and report.json / JUnit / mp4 evidence.

This repository is the MIT-licensed agent layer:

  • mcp/ — MCP stdio server + CLI + session/capture engine (also on npm as bezlo-mcp)
  • action.yml + examples/ — composite GitHub Action (PR screenshots + JUnit + video)
  • skill/ — the bezlo-preview skill for Claude Code / skills-compatible agents
  • server.json — official MCP registry manifest
  • docs/MCP.md — full tool reference

The Chrome extension itself is proprietary and lives in a separate private repo (extension/, store art, brand). This repository contains no extension code, store assets or brand artwork. The name Bezlo and its logo are not licensed for reuse.

Install

# MCP (Claude Code / Claude Desktop / Cursor)
npm install -g bezlo-mcp
claude mcp add bezlo -- npx -y bezlo-mcp

# or run from a checkout
node mcp/cli.mjs --list

The package bundles no browser: install Chrome/Chromium (or set RP_CHROME). Chrome for Testing is recommended; WebKit mode additionally needs npx playwright install webkit.

Quick CLI

node mcp/cli.mjs --url http://localhost:3000 --device iphone16 --out shot.png
node mcp/cli.mjs --url http://localhost:3000 --matrix --devices iphone16,ipadpro11,px10 --out ./matrix
node mcp/cli.mjs --url http://localhost:3000 \
  --actions '[{"type":"click","text":"Sign in"},{"type":"type","selector":"#email","value":"[email protected]","submit":true}]' \
  --record --assert '[{"type":"noOverflow"},{"type":"noConsoleErrors"}]' --out after.png

GitHub Action

- uses: gba3124/bezlo@main
  id: bezlo
  with:
    url: http://localhost:3000
    device: iphone16
    steps: '[{"type":"click","text":"Sign in"}]'
    asserts: '[{"type":"noOverflow"},{"type":"noConsoleErrors"}]'
    record: 'true'

Artifacts: final screenshot, report.json, junit.xml, optional mp4/gif. Outputs: pass, report, junit, video, gif, screenshot, artifacts.

Engines

  • chromium (default) — Chrome's own emulation engine (same as the extension's frames): real touch, mobile UA and client hints, DPR; full recording and CDP accessibility.
  • webkit — Playwright WebKit, a Safari-flavoured approximation (not real iOS Safari; no recording / CDP a11y).

License

MIT — see LICENSE. The Chrome extension, store assets and the Bezlo brand are not covered by this license and are not part of this repository.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.