Dock AI

What the Dock AI MCP server does

The Dock AI MCP server helps an agent move from a real-world business to an MCP endpoint that may support actions for that business. It queries the Dock AI registry, which maps businesses to MCP connectors. The input is a business domain such as example-restaurant.com; the response can identify one or more matching entities and available providers.

Returned entity data can include the business name, path, city, and country. For each matching connector, the response may provide the provider name, MCP endpoint, entity identifier, supported capabilities, and verification information. If a business has a known provider without a published connector, the registry can also report pending providers.

Typical agent flow:

1. Find the official domain for a business.
2. Call resolve_domain with that domain.
3. Inspect the returned connector and its capabilities.
4. Connect to the returned MCP endpoint when an appropriate connector exists.

How it works

The resolve_domain tool accepts one required string input: domain. The server sends that domain to the Dock AI registry API and returns the registry result. A successful result can include a claude_desktop_config object showing how to add a discovered provider to Claude Desktop.

The Dock AI MCP server does not itself perform reservations, availability checks, or other business operations. Instead, it helps an agent discover the MCP server that may provide those operations. The examples describe using discovered connectors for restaurant reservations and hotel bookings, but the available actions depend on the provider returned for the requested entity.

Setup and configuration

The hosted service is available at https://connect.dockai.co/mcp, so no local package installation is required. Add that URL as an MCP server in a compatible client:

{
  "mcpServers": {
    "dock-ai": {
      "url": "https://connect.dockai.co/mcp"
    }
  }
}

For self-hosting, the repository provides a Vercel deployment option. To run locally, use uvx dock-ai-mcp, or install the package with pip install dock-ai-mcp and start it with dock-ai-mcp. The local process listens on http://0.0.0.0:8080/mcp.

No required environment variables are documented. The server queries the Dock AI registry API and states that it does not collect or store user data.

Tools and capabilities

resolve_domain

Resolves a business domain against the registry. Its response may include:

• Matching business entities and locations
• MCP provider names and endpoints
• Provider-specific entity IDs
• Declared capabilities such as reservations or availability
• Connector verification level and method
• A Claude Desktop configuration block
• Pending providers that do not yet have an MCP connector

The tool is suited to agents that first identify a business website and then need to discover whether an MCP connection exists for that business.

Limitations and notes

Results depend on the Dock AI registry and the domain supplied by the caller. A business may have no connector, or its provider may be listed as pending until an MCP connector is published. Discovery does not guarantee that every returned provider supports the action an agent wants to perform.

The server is a discovery layer rather than a general booking or commerce interface. After resolution, the agent must connect to and use the discovered provider’s MCP server. The project documents MIT licensing, support through GitHub Issues, and additional documentation at dockai.co/docs.