MCP server

Waveform exposes its entire engine over the Model Context Protocol — the open standard AI clients like Cursor and Claude Desktop use to talk to external tools. Everything the web chat can do, your editor's AI can do: create projects, build pages, edit content, deploy, configure domains, run audits.

Who this is for

Developers and agencies who live in their editor. Build a client site from Cursor while looking at the client's requirements doc; script repetitive site operations through Claude Desktop; wire Waveform into an agent workflow you already run.

Connecting

The MCP server speaks streamable HTTP and authenticates with your Waveform access token as a Bearer token.

In an MCP-capable client, add a server entry pointing at the Waveform MCP endpoint with an Authorization: Bearer <your-token> header. For Cursor, that looks like:

{
  "mcpServers": {
    "waveform": {
      "url": "https://mcp.waveform.adwave-apps.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_WAVEFORM_TOKEN"
      }
    }
  }
}

Your MCP client will list the available tools once connected.

What you can do

The full registry — 150+ tools across 33 domains — is available, including:

• Projects — create, list, open, delete
• Building — pages, sections, content, themes, images
• Design pipeline — build from a brief, import from URL or image
• Quality — audits (SEO, accessibility, copy, links) with auto-fix
• Publishing — deploy, promote, rollback, domains
• Content — collections, entries, CMS adapters
• Growth — forms, analytics, experiments, webhooks
• Safety — history, undo, checkpoints

A tool_search meta-tool helps clients discover advanced tools on demand, so the common tool surface stays small and fast.

Same engine, same project

The MCP server and the web app operate on the same projects. Start a site in chat during a call, refine it from Cursor in the afternoon, and hand the client the web dashboard for content edits — it's all one project.