DMCA Takedown MCP Server

The first DMCA takedown & host-reveal MCP server. Connect it to Claude (or any MCP client) and your AI agent can answer: “who really hosts this infringing site — even behind Cloudflare — where does the DMCA/abuse notice go, and how do I actually get it taken down?”

Hosted and operated by Lawyerd, a counsel-led content-removal practice. This repository is the public home and documentation for the server; it is a remote, hosted MCP server — there is nothing to install or run yourself. Just point your agent at the endpoint.

• Endpoint: https://mcp.lawyerd.net/takedown (Streamable HTTP)
• Alias: https://mcp.lawyerd.net/mcp
• Registry: net.lawyerd/takedown in the official MCP Registry
• Cost: free, no account, rate-limited
• Discovery: /.well-known/mcp.json

Quick start

Add the server to Claude Code / Claude Desktop:

claude mcp add --transport http lawyerd https://mcp.lawyerd.net/takedown

Or add it to any MCP client config (mcpServers):

{
  "mcpServers": {
    "lawyerd": {
      "type": "http",
      "url": "https://mcp.lawyerd.net/takedown"
    }
  }
}

Then just ask your assistant, in your own words:

“Someone reposted my photo set and is selling prints at https://example-infringer.com. Who hosts it and how do I take it down?”

The agent chains the three tools below and hands you a plain-language answer.

Tools

resolve_host

Look up who hosts a URL and where a DMCA/abuse notice would go. Identifies the platform, the CDN/proxy in front (e.g. Cloudflare, DDoS-Guard), or the direct hosting provider and its abuse contact. For domains whose real host has already been revealed from behind a proxy, it returns that host immediately; for a domain still hidden behind a proxy it says so honestly — revealing the true host means initiating an abuse report. Read-only; files nothing.

assess_takedown

The plan for a takedown of a specific URL: the route it would follow (host abuse, platform form, or reveal-first), what the rights holder must supply (proof of ownership and a one-time signed attestation), the step-by-step process, and the legal caveats — including 17 U.S.C. §512(f) and scope limits. Read-only; it does not judge the merits of the claim.

start_takedown

Returns a secure onboarding link to actually begin a takedown. A human opens it once in a browser to identify themselves, sign the authorization (the one step that legally requires a human), and provide proof of ownership. Nothing is filed by the agent.

Example

// tools/call → resolve_host  { "url": "https://example-infringer.com" }
{
  "domain": "example-infringer.com",
  "frontedBy": "cdn_proxy",
  "proxy": { "name": "Cloudflare" },
  "realHostHidden": true,
  "takedownRoute": "needs_reveal",
  "nextStep": "The site is behind Cloudflare — the real host is hidden. Revealing it requires initiating an abuse report; start a takedown to do that."
}

For a domain already revealed, resolve_host returns the real host and its abuse email directly, with revealSource: "cached_reveal".

What this is — and is not

• ✅ A lookup and planning layer for DMCA / abuse takedowns, callable by AI agents.
• ✅ Honest about what it knows: it never invents an abuse contact, and it does not claim a synchronous “reveal behind Cloudflare” it cannot actually perform without filing.
• ❌ Not an auto-filing bot. A DMCA notice is signed by the rights holder under penalty of perjury; that signature is always made by a human.
• ❌ Not legal advice, and it does not guarantee removal.
• ❌ Not a tool for suppressing lawful content. Copyright, trademark and brand-impersonation matters only — lawful criticism, reviews and journalis

... View full README on GitHub