DevLab MCP Suite

DevLab MCP Suite

Modular testing and automation MCP servers for Android devices, desktop browsers, canvas games, and visual regression.

Category
Visit Server

README

DevLab MCP Suite

Modular testing and automation MCP servers for Android devices, desktop browsers, canvas games, and visual regression.

DevLab MCP Suite
├── @tanguito/android-dev-mcp        ← Android device automation (ADB)
├── @tanguito/browser-dev-mcp        ← Desktop browser automation (Playwright)
├── @tanguito/visual-regression-mcp  ← Visual regression & pixel diff
└── @tanguito/devlab-shared          ← Shared contracts and helpers

The Testing Triangle

             ┌──────────────────────┐
             │   visual-regression   │  ← pixel-perfect diffs
             │   screenshot compare  │
             └──────────┬───────────┘
                        │
         ┌──────────────┴──────────────┐
         │                             │
  ┌──────┴──────┐              ┌───────┴──────┐
  │  android     │              │   browser     │
  │  device tests│              │   canvas/web  │
  └──────────────┘              └──────────────┘

browser-dev-mcp screenshot  →  visual-regression-mcp diff  →  markdown report

Three independent testing layers, same tool patterns, same evidence system, single monorepo.

5-Minute Quick Start

# 1. Clone and install
git clone https://github.com/Tanguito86/devlab-mcp.git
cd devlab-mcp
pnpm install

# 2. Environment check + auto-fix
pnpm setup

# 3. Build and test
pnpm build
pnpm test          # 67 tests: 42 android + 8 browser + 8 visual + 9 shared

# 4. Your first browser workflow
node examples/browser-hello-world/run-hello-world.js

You just ran a 5-step browser workflow: open → screenshot → console → errors → report. Takes ~5 seconds.

Package Table

Package Purpose Version Tests npm Status
@tanguito/devlab-shared Types, schemas, helpers 0.1.0 9 @tanguito/devlab-shared Pending publish
@tanguito/android-dev-mcp Device automation (ADB) 1.2.0 42 @tanguito/android-dev-mcp ✅ Published
@tanguito/browser-dev-mcp Browser/canvas (Playwright) 1.0.0 8 @tanguito/browser-dev-mcp Pending publish
@tanguito/visual-regression-mcp Screenshot diffing 0.1.0 8 @tanguito/visual-regression-mcp Pending publish

Total: 67 tests | 70 tools | 4 packages

Publishing: android-dev-mcp is already on npm. shared, browser, and visual-regression are ready. See publishing guide for exact commands.

Case Study: Galaxy Raiders

We validated browser-dev-mcp and visual-regression-mcp against a real 68-script HTML5 Canvas shmup with 5 bosses across 20 levels.

Browser Testing

Metric Result
Workflows executed 4 (smoke, boss ladder, performance, console audit)
Total steps 81/81 (100%)
Page errors 0 across all workflows
Bosses validated 5/5 (Crabtron, Serpentrix, Colossus, Lieutenant, Emperor)
Screenshots 13 real rendered canvas captures
Leak test 3/3 consecutive cycles clean

Visual Regression

boss-lv5:   81 kB, 360×640, 0 changed pixels, ✅ PASS
boss-lv20:  90 kB, 360×640, 0 changed pixels, ✅ PASS

Pipeline: browser-dev-mcp captures canvas screenshots → visual-regression-mcp diffs against baselines → markdown report with pass/fail.

Read the full case study →

Packages

browser-dev-mcp

Desktop browser automation via Playwright. 24 tools for navigation, screenshots, canvas game testing, JavaScript evaluation, keyboard/mouse input, and evidence capture.

pnpm --filter @tanguito/browser-dev-mcp build
pnpm --filter @tanguito/browser-dev-mcp test
pnpm --filter @tanguito/browser-dev-mcp run doctor

Key tools: browser_open_url, browser_screenshot, browser_screenshot_canvas, browser_click, browser_click_percent, browser_press_key, browser_type_text, browser_evaluate_js, browser_evaluate_game_state, browser_wait_for_canvas_change, browser_get_console_logs, browser_get_page_errors, and 12 more.

Full README →

visual-regression-mcp

Visual regression testing — pixel comparison, diff generation, markdown reporting. 4 tools: visual_compare_images, visual_create_baseline, visual_compare_folder, visual_generate_report. Zero native dependencies — pure Node.js PNG parsing with zlib.

pnpm --filter @tanguito/visual-regression-mcp build
pnpm --filter @tanguito/visual-regression-mcp test
pnpm --filter @tanguito/visual-regression-mcp run doctor

Full README →

android-dev-mcp

Android device automation via ADB. 42 tools for device inspection, app management, UI interaction, screenshots, logcat, and workflow execution.

pnpm --filter @tanguito/android-dev-mcp build
pnpm --filter @tanguito/android-dev-mcp test

Full README →

devlab-shared

Minimal shared contracts: textResponse, sanitizeName, validateSessionId, RegisterTool, WorkflowStep, StepResult, and base evidence types. Zero IO, zero runtime dependencies. 9 tests, full doctor check. 107 lines.

pnpm --filter @tanguito/devlab-shared build
pnpm --filter @tanguito/devlab-shared test
pnpm --filter @tanguito/devlab-shared run doctor

Full README →

Examples

Example Time What it does
browser-hello-world ~5s Opens test page, takes screenshot, collects console/errors
galaxy-raiders ~15s Loads Galaxy Raiders, starts game, captures canvas, jumps to boss
galaxy-visual-regression ~2s Pixel-diff Galaxy Raiders boss screenshots against baselines
# Hello World
node examples/browser-hello-world/run-hello-world.js

# Galaxy Raiders (requires game running on localhost:5173)
node examples/galaxy-raiders/run-galaxy-smoke.js

Scripts

Command Description
pnpm setup Detect environment, auto-install Chromium, suggest fixes
pnpm build Build all packages
pnpm test Run all 67 tests
pnpm doctor Health checks per package (use pnpm --filter <pkg> run doctor)
pnpm typecheck TypeScript type checking
pnpm clean Remove dist/ from all packages
pnpm pack:dry-run Preview npm package contents
pnpm changeset Create a version changeset
pnpm changeset:version Bump versions from changesets
pnpm changeset:publish Publish changed packages to npm

Troubleshooting

"Browser not open" on WSL

WSL's headless Chromium shell is missing shared libraries. browser-dev-mcp automatically uses the full Chromium binary if PLAYWRIGHT_CHROMIUM_EXECUTABLE_PATH is set.

# Find your Chromium binary
ls ~/.cache/ms-playwright/chromium-*/chrome-linux64/chrome

# Set the env var
export PLAYWRIGHT_CHROMIUM_EXECUTABLE_PATH=/home/user/.cache/ms-playwright/chromium-1223/chrome-linux64/chrome

Run pnpm setup — it detects WSL and suggests the fix automatically.

Playwright not found

pnpm setup       # Auto-detects and suggests fixes
npx playwright install chromium   # Manual install

ADB not found (android-dev-mcp only)

ADB is optional — only needed for @tanguito/android-dev-mcp.

# Install Android SDK Platform Tools
# https://developer.android.com/tools/releases/platform-tools

Repository Structure

devlab-mcp/
  pnpm-workspace.yaml
  package.json              # Root scripts + pnpm setup
  .changeset/               # Changesets configuration
  .github/workflows/ci.yml  # CI matrix (Node 20/22)
  scripts/
    devlab-setup.js         # Environment detector + auto-fixer
  packages/
    shared/                    # @tanguito/devlab-shared (107 lines)
    android-dev-mcp/           # @tanguito/android-dev-mcp (42 tools)
    browser-dev-mcp/           # @tanguito/browser-dev-mcp (24 tools)
    visual-regression-mcp/     # @tanguito/visual-regression-mcp (4 tools)
  examples/
    browser-hello-world/       # 5-step first workflow
    galaxy-raiders/            # Real canvas game smoke test
    galaxy-visual-regression/  # Pixel-diff boss screenshots
  docs/
    devlab-suite-v1-freeze.md      # v1 freeze document
    architecture.md                # Suite architecture decisions
    case-study-galaxy-raiders.md   # Full Galaxy Raiders validation
    publishing.md                  # How to version and publish

Architecture

  • Monorepo with pnpm workspaces — shared iteration, single CI
  • Independent packages — each installable, runnable, testable in isolation
  • Shared is minimal — 107 lines of pure contracts, zero runtime coupling
  • Changesets for independent versioning within the monorepo
  • Workflow-first — declarative JSON test plans executed by MCP servers

Full architecture → Freeze document →

License

MIT — see individual package LICENSE files.

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured
Exa Search

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.

Official
Featured