CloakBrowser MCP
About
CloakBrowser MCP server for AI agents: Playwright-powered browsing, clean tool forwarding, Docker support, and multi-session HTTP transport.
Details
- Author
- swimmwatch
- Categories
- Web Scraping, Automation, Other
Jump to
Setup
Install CloakBrowser MCP in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/swimmwatch/cloakbrowser-mcp
Follow the installation instructions in the repository README, then restart your MCP client.
CloakBrowser MCP server for AI agents: Playwright-powered browsing, clean tool forwarding, Docker support, and multi-session HTTP transport.
cloakbrowser-mcpis a drop-in Playwright MCP-compatible browser automation server with unchanged upstream tools,CloakBrowserChromium, and production-ready npm, Docker, and Streamable HTTP packaging. It runs upstream@playwright/mcpas the canonical tool surface and points that runtime at CloakBrowser.
Runnpx -y cloakbrowser-mcp@latest, connect Claude Desktop or Codex CLI, ask for web research, daily automation, or testing in plain English, and inspect the real browser result.
Documentation:swimmwatch.github.io/cloakbrowser-mcp·Comparison·Recipes
- Playwright MCP browser automation backed by CloakBrowser;
- unchanged upstream browser tools plus two local introspection tools;
- npm or Docker installation over stdio or Streamable HTTP;
- persistent browser profiles, validated context options, and Chrome extension loading;
- GeoIP-aware proxy matching for regional QA;
- humanized mouse, keyboard, and scroll behavior for interaction-sensitive flows.
Cross-platform checks cover npm on Linux x64/arm64, macOS arm64/x64, and Windows x64 across Node.js 22 and 24-26. Docker images are built and smoke-tested forlinux/amd64andlinux/arm64.
See@playwright/mcpvscloakbrowser-mcpwhen deciding whether plain upstream Playwright MCP or CloakBrowser MCP fits a deployment better. TheRecipespages show task-focused setup paths for persistent login profiles, Chrome extensions, reverse proxies, regional QA, client connections, and CI smoke tests.
Requires Node.js 22.13+ in the 22.x line, or Node.js 24+. Run diagnostics before wiring a client:
npx -y cloakbrowser-mcp@latest --transport streamable-http --http-port 3000
See the generatedCLI Referencefor all flags.
docker run --rm --init -i \ -v "$PWD/artifacts:/data" \ swimmwatch/cloakbrowser-mcp:latest
docker run --rm --init -p 127.0.0.1:3000:3000 \ -v "$PWD/artifacts:/data" \ swimmwatch/cloakbrowser-mcp:latest \ --transport streamable-http --http-host 0.0.0.0 --http-port 3000
The Docker image writes artifacts to/dataand is published forlinux/amd64andlinux/arm64. It defaults toCLOAK_PLAYWRIGHT_MCP_NO_SANDBOX=truefor compatibility with containerized runtimes where Chromium sandboxing is often unavailable. If your host and container runtime support Chromium sandboxing, setCLOAK_PLAYWRIGHT_MCP_NO_SANDBOX=false; for untrusted pages, keep container network access and mounted host directories tightly scoped. The same tags are also available fromghcr.io/swimmwatch/cloakbrowser-mcp. SeeDockerfor persistent profiles, extension mounts, HTTPS, and smoke-test examples, or use thereverse proxy recipefor a focused Streamable HTTP deployment.
codex mcp add cloakbrowser -- npx -y cloakbrowser-mcp@latest
claude mcp add --transport stdio cloakbrowser -- npx -y cloakbrowser-mcp@latest
{ "servers": { "cloakbrowser": { "type": "stdio", "command": "npx", "args": ["-y", "cloakbrowser-mcp@latest"] } } }
Claude Desktop, Cursor, Cline, Windsurf, Warp, And OthermcpServersClients
Add this server entry to the client's MCP JSON config:
{ "mcpServers": { "cloakbrowser": { "command": "npx", "args": ["-y", "cloakbrowser-mcp@latest"] } } }
{ "mcpServers": { "cloakbrowser": { "command": "docker", "args": [ "run", "--rm", "--init", "-i", "-v", "/tmp/cloakbrowser-artifacts:/data", "swimmwatch/cloakbrowser-mcp:latest" ] } } }
npx -y cloakbrowser-mcp@latest --transport streamable-http --http-port 3000 codex mcp add cloakbrowser --url http://127.0.0.1:3000/mcp claude mcp add --transport http cloakbrowser http://127.0.0.1:3000/mcp
Paste this into Codex, Claude Code, Copilot, Cursor, Cline, Windsurf, or a similar coding assistant that can edit MCP config:
Install the CloakBrowser MCP server for this workspace. Name it "cloakbrowser". Prefer stdio with command "npx" and args ["-y", "cloakbrowser-mcp@latest"]. If this client uses VS Code mcp.json, add it under "servers" with type "stdio". If this client uses Claude/Cursor/Cline/Windsurf/Warp-style config, add it under "mcpServers" with the same command and args. Do not add secrets.
More examples are inGetting Started, with dedicated recipes forClaude DesktopandCodex CLI.
Use upstreamPLAYWRIGHT_MCP_variables for browser, artifacts, timeouts, network, and tool capability settings. Cloak-specific bridge toggles useCLOAK_PLAYWRIGHT_MCP_. Select a Pro Preview browser build before startup with--release-channel previeworCLOAK_PLAYWRIGHT_MCP_RELEASE_CHANNEL=preview; the default isstable.
The common variable table now lives inConfiguration. That page also covers persistent profiles, validated context options, Chrome extensions, Streamable HTTP metadata, and HTTPS/auth options. SeeGeoIP Proxy Matchingfor regional proxy behavior,Humanized Input Behaviorfor interaction realism, andRecipesfor task-focused configurations.
SeeVersion Compatibilityfor the maintained compatibility table.
The upstream Playwright MCP tool list is authoritative. This project does not reimplement or re-document upstream browser schemas in source code.
- cloakbrowser_binary_inforeturns CloakBrowser package, platform, cache, and resolved binary data.
- cloakbrowser_bridge_inforeturns bridge metadata, upstream package/version, and local tool names.
npm install npm run build npm test npm run docker:build npm run docker:smoke npm run server:validate npm run bridge:compare -- cloakbrowser-mcp:dev --report bridge-parity-report.json
Documentation starts atdocs/getting-started.md. Contributor material is grouped underdocs/contributor-guide.md.
A browser screenshot tool to capture scrolling screenshots of webpages using Playwright, with support for intelligent section identification and multiple output formats.
Take screenshots and read console logs from web pages using Playwright.
Control a browser for web automation tasks using Playwright on Cloudflare Workers.
Control a browser for web automation tasks like navigation, typing, clicking, and taking screenshots using Playwright on Cloudflare Workers.
Fetch and extract web content using a Playwright headless browser, with support for intelligent extraction and flexible output.
Query financial web tables from sources like iwencai, tdx, and eastmoney using Playwright.
A production-ready web scraping platform with ML-powered automation, browser automation via Playwright, and persistent caching.
Control a browser for automation and web scraping tasks using Playwright.
A server providing Playwright tools for browser automation and web scraping.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
