passiv-mcp

# passiv-mcp

MCP server that gives AI agents free web tools — zero auth, zero dependencies.

Four tools, backed by the free [Passiv](https://hermes-passiv.pages.dev) APIs:

| Tool | What it does |
|------|--------------|
| `html_to_markdown` | Convert HTML to clean Markdown or plain text (max 50 KB/call) |
| `compliance_scan` | 9-point EU compliance scan: privacy policy, terms, cookie consent, imprint, accessibility statement, DPA, security headers, meta tags, hreflang |
| `profile_page` | Page profile: meta/OG tags, JSON-LD, heading structure, alt-text coverage, security headers, score + grade |
| `check_security_headers` | CSP, HSTS, X-Frame-Options, X-Content-Type-Options, Referrer-Policy |

Listed in the official [MCP Registry](https://registry.modelcontextprotocol.io) as `io.github.mahope/passiv-mcp`.

Works with any MCP client: Claude Desktop, Claude Code, Cursor, Windsurf, and others.

## Install

No dependencies. Node.js 18+ required.

### Run from GitHub (no clone needed)

```bash
npx github:mahope/passiv-mcp
```

### Claude Desktop

Add to `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "passiv": {
      "command": "npx",
      "args": ["github:mahope/passiv-mcp"]
    }
  }
}
```

### Claude Code

```bash
claude mcp add passiv -- npx github:mahope/passiv-mcp
```

## Example prompts

Once connected, try:

- "Convert this HTML snippet to Markdown: <h1>…</h1>…"
- "Scan example.com for GDPR compliance basics"
- "Check the security headers on my site"
- "Profile this page and tell me what meta tags are missing"

## Privacy

URLs sent to a tool are fetched server-side by the Passiv scan proxy and not stored.
HTML sent to `html_to_markdown` is converted and discarded immediately.

## Related

- [Clean Copy](https://github.com/mahope/clean-copy) — browser extension, same Markdown engine
- [clean-copy-cli](https://github.com/mahope/clean-copy-cli) — CLI version
- [compliance-site-check](https://github.com/mahope/compliance-site-check) — GitHub Action with the compliance engine

## License

MIT