Delv
CommunityAbandoned· 11mo4.3by Anand Tyagi

Webpage Screenshot MCP

Captures full-page and element-specific screenshots for UI development feedback, with authentication and system browser support.

C
Safety & Trust

Delv Safety Grade: C

Score 54/100 · assessed 2026-04-28

Maintainer40
Permissions45
Supply chain35
Transparency70
Incidents100

Webpage Screenshot MCP is a community tool by solo developer Anand Tyagi for capturing browser screenshots. The server launches system browsers with full desktop control capabilities, including authentication handling. Whilst the repository is open source with reasonable documentation, the maintainer appears to be a solo developer with limited track record. The supply chain is weak: no npm package exists, requiring manual clone-and-build installation. The permissions model is concerning as it grants broad browser control and network access without sandboxing. Desktop browser automation inherently carries risk of credential exposure if misused. No security incidents are known, but the combination of solo maintenance, manual install, and powerful browser control warrants caution. Suitable for development environments but requires careful consideration for production use.

Lethal Trifecta (prompt-injection exposure)

TWO OF THREE
Private dataNo
Reads secrets, credentials, private files
Untrusted inputYes
Ingests web pages, PRs, issues, emails
External commsYes
Can send data outbound

Screenshot tools render arbitrary URLs and return image content. Same axes.

Green flags

  • Open source repository with visible code
  • Clear documentation of screenshot capabilities
  • No known security incidents or malicious behaviour
  • Specific use case (UI development) limits scope

Red flags

  • Solo maintainer with limited public track record
  • No package distribution, requires manual clone and build
  • Full system browser control with authentication handling
  • Broad desktop automation capabilities without sandboxing
  • Could access credentials stored in system browser

Permissions requested

Browser controlOutbound networkWrite filesDesktop control
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

npm install && npm run build

Review

Webpage Screenshot MCP does exactly what the name suggests: it grabs full-page screenshots or captures specific DOM elements from any URL you throw at it. It's built on Puppeteer, so it handles JavaScript-heavy sites properly and can authenticate behind login walls if you pass cookies. The system browser support means you can use your local Chrome or Edge profile, which is handy when you're testing authenticated flows without faffing about with session tokens. I'd reach for this when I'm building UI components and need to verify how they render across different viewport sizes or when I'm documenting bugs for a designer. The element-specific screenshot feature is genuinely useful here: you can target a CSS selector and get just that component, not the entire page with scrollbars and browser chrome. It's faster than manually cropping screenshots and more reliable than asking a designer to squint at a staging URL. The pre-listed use cases mention visual regression testing, and that's fair, though you'll need to wire up your own diffing logic. This tool just captures the images. It doesn't compare them or flag changes. That's not a criticism, just a reality check. If you're expecting an all-in-one regression suite, you'll be disappointed. If you want a reliable way to grab screenshots from Claude Desktop or Cursor and pipe them into your own workflow, this does the job. Quirks: it's a community project, so documentation is thin. The repo README covers the basics but doesn't walk through authentication flows or edge cases. I had to read the source to figure out how to pass custom headers. Also, the build step is mandatory, which adds friction if you're used to plug-and-play MCP servers. No pre-built binaries, no npm package. You clone the repo, run npm install, build it, then point your host config at the compiled output. Who shouldn't bother: if you're not doing UI work or you don't need programmatic screenshots, skip it. If you're already using a dedicated visual regression tool like Percy or Chromatic, this won't replace it. But if you're a solo dev or small team who needs screenshots on demand inside Claude, it's a solid addition to your toolkit.
Verdict

Install this if you're doing UI development or QA and want Claude to grab screenshots without leaving the conversation. Skip it if you need a full regression testing suite or if the manual build process sounds like too much hassle. It's a focused tool that does one thing well, assuming you're comfortable with a bit of setup.

Good at

  • Handles JavaScript-heavy sites properly via Puppeteer, so dynamic content renders correctly.
  • Element-specific screenshots via CSS selectors save time compared to cropping full-page captures manually.
  • System browser support means you can reuse your local Chrome profile for authenticated flows.
  • Works across Claude Desktop, Claude Code, and Cursor without platform-specific quirks.

Watch out

  • No npm package or pre-built binaries, so you're cloning and building from source every time.
  • Documentation is sparse: authentication flows and advanced options require reading the code.
  • Doesn't include diffing or comparison logic, so you'll need to build your own regression workflow around it.
  • The build step adds friction compared to plug-and-play MCP servers.

Use cases

  • visual regression testing
  • UI QA
  • design feedback loops
  • agent verification

Getting started

1. Clone the repo from https://github.com/ananddtyagi/webpage-screenshot-mcp and run `npm install && npm run build` in the project directory. 2. Add the server to your Claude Desktop or Cursor config, pointing the command at the built index.js file in the build folder. 3. Restart your host and ask Claude to take a screenshot of a public webpage to verify it's working. 4. For authenticated sites, you'll need to pass cookies or use the system browser flag, which requires reading the source to understand the exact syntax. 5. Watch out for the lack of pre-built binaries: every update means rebuilding from source.

Works with

Claude DesktopClaude CodeCursor

Similar MCPs