SuperAPI AI Directory

web-exposure-mcp

Read-only check of a live URL for publicly exposed secret files: .git, .env, JS source maps, backup/SQL dumps, directory listings and dotfiles.

Open website

About this tool

Purpose

web-exposure-mcp is an MCP (Model Context Protocol) server used to let AI agents confirm, against a deployed online URL, whether sensitive files are actually publicly served. The targets it checks include: exposed .git, .env secrets, JavaScript source maps, backup/SQL dumps, directory listings, and dotfiles.

Unlike other tools that only give a list of "possible issues," this tool actually fetches bytes and verifies content, reporting only items that are truly reachable, along with evidence.

Main capabilities

  • Probes whether an online URL publicly serves sensitive files, covering .git, .env, JS source maps, backup/SQL dumps, directory listings, and dotfiles.
  • Avoids false positives by reading bytes and fingerprinting content. For example: .git/config must parse as a git config, .env must contain secret lines in KEY=VALUE form, and archive files must begin with real magic bytes.
  • Each check is triggered at most once, and is reported only when the provided bytes can prove it.
  • Read-only: the scanner never writes anything to the target, does not follow redirects to other hosts, and reads at most 64 KB per file (so it can fingerprint multi-GB backups without downloading them).
  • Output is JSON on stdout (pipeable into CI), with a one-line summary on stderr.

Usage

In Claude Desktop, Cursor, or any MCP client, add it to the mcpServers configuration. Then instruct the agent, for example: "Scan https://staging.myapp.com for publicly exposed secret files."

Run in one line, no installation, no API key.

Optional installation: zero dependencies, pure Node ≥18. Every request goes directly from the tool to the target you specify, and no data leaves your machine.

Use cases

  • Conduct external exposure surface audits on deployed online environments, examining the deployment edge from the perspective actually visible to an attacker.
  • Integrate into CI for automated checks via JSON output on stdout.
  • Need to confirm whether sensitive files such as .git and .env are truly publicly accessible, rather than relying only on a checklist.

Limitations

  • Read-only, does not write to the target, does not follow redirects to other hosts.
  • Reads at most 64 KB per file.
  • Requires Node ≥18.
  • No other platform support, billing model, or pricing information is provided in the materials (apart from the fixed-scope external exposure audit service at $99 / 24h, nothing else is currently provided).