For AI Agents

Use the Dock AI MCP server to discover connectors for real-world businesses.

Quick Start

Add Dock AI to your Claude Desktop configuration:

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

Config file location: ~/Library/Application Support/Claude/claude_desktop_config.json

Remote URL: https://connect.dockai.co/mcp

No installation required. Just add the URL to your MCP client.

How it works

User asks to interact with a business (e.g., "Book a table at Septime")
AI calls resolve_domain("septime-charonne.fr")
Dock AI returns the entity + available connectors + capabilities
AI calls execute_action with the required parameters

Authentication

The MCP server uses OAuth 2.0 to authenticate users. When you add the Dock AI connector to your AI assistant, you'll be prompted to sign in once.

How OAuth works with MCP

Add the Dock AI connector to Claude, ChatGPT, or Mistral
First time you use it, you'll see a "Connect" button
Sign in with your email (6-digit code, no password needed)
Done! The connector handles authentication automatically

Why authentication? Actions like booking or sending messages are executed on your behalf. Authentication ensures traceability and prevents abuse.

For direct API access

If you need to call the API directly (not via MCP), request an API key at support@dockai.co.

Tool: resolve_domain

Resolve a domain to its entity and connectors.

Parameter	Type	Description
domain	string	Domain to resolve (e.g., "septime-charonne.fr")

Response

{
  "domain": "septime-charonne.fr",
  "entity": {
    "id": "uuid",
    "name": "Septime",
    "domain": "septime-charonne.fr",
    "category": "restaurant",
    "city": "Paris",
    "country": "FR"
  },
  "connectors": [
    {
      "slug": "zenchef",
      "name": "Zenchef",
      "category": "booking",
      "status": "connected",
      "mcp_endpoint": "https://mcp.zenchef.com",
      "external_id": "12345"
    }
  ],
  "capabilities": [
    {
      "action": "book",
      "name": "Booking",
      "description": "Book a table. Open Tuesday to Saturday.",
      "input_schema": {
        "date": { "type": "string", "required": true, "description": "Date", "format": "date" },
        "time": { "type": "string", "required": true, "description": "Time", "format": "time" },
        "guests": { "type": "number", "required": true, "description": "Number of guests" },
        "name": { "type": "string", "required": true, "description": "Name" }
      }
    },
    {
      "action": "send_message",
      "name": "Contact",
      "description": "Send a message to the team.",
      "input_schema": {
        "name": { "type": "string", "required": true, "description": "Name" },
        "email": { "type": "string", "required": true, "description": "Email", "format": "email" },
        "message": { "type": "string", "required": true, "description": "Message" }
      }
    }
  ],
  "_ai_hint": "MCP endpoint found for Septime! ..."
}

Response Fields

Field	Description
entity	Business info (name, location, category)
connectors	List of software providers for this business
capabilities	Actions available for this business (booking, contact, etc.)
capabilities[].action	Action identifier (book, send_message, search_catalog...)
capabilities[].input_schema	Parameters required for this action (type, required, description, format)
connectors[].status	"connected" = MCP available, "not_connected" = detected but no MCP yet
connectors[].mcp_endpoint	MCP server URL (only if connected)
_ai_hint	Human-readable hint for the AI assistant

Tool: execute_action

Execute a business action discovered via resolve_domain.

Requires authentication. The MCP handles this automatically. See Authentication above.

Parameter	Type	Description
entity_id	string	Entity UUID from resolve_domain response
action	string	Action identifier (e.g., "book", "send_message")
params	object	Parameters matching the action's input_schema

Response

{
  "success": true,
  "action": "book",
  "result": {
    "confirmation": "Your booking has been confirmed",
    "reference": "RES-12345"
  }
}

Usage Examples

Example 1: Restaurant Booking via Capabilities

// User: "Book a table at Septime Paris for 4 people on Friday"

// Step 1: AI searches web for restaurant domain
// Found: septime-charonne.fr

// Step 2: AI calls resolve_domain
resolve_domain("septime-charonne.fr")

// Step 3: Response includes capabilities
{
  "entity": { "id": "abc-123", "name": "Septime", "city": "Paris" },
  "capabilities": [
    {
      "action": "book",
      "name": "Booking",
      "description": "Book a table. Open Tuesday to Saturday.",
      "input_schema": {
        "date": { "type": "string", "required": true, "format": "date" },
        "time": { "type": "string", "required": true, "format": "time" },
        "guests": { "type": "number", "required": true },
        "name": { "type": "string", "required": true }
      }
    }
  ]
}

// Step 4: AI calls execute_action with the parameters
execute_action("abc-123", "book", {
  "date": "2025-02-14",
  "time": "20:00",
  "guests": 4,
  "name": "John Smith"
})

// Step 5: Booking confirmed!
{ "success": true, "result": { "confirmation": "Booking confirmed" } }

Example 2: No MCP Available

// User: "Order from Pizza Hut"
resolve_domain("pizzahut.fr")

// Response
{
  "entity": { "name": "Pizza Hut France", "category": "restaurant" },
  "connectors": [
    {
      "slug": "deliveroo",
      "name": "Deliveroo",
      "status": "not_connected"
    },
    {
      "slug": "ubereats",
      "name": "Uber Eats",
      "status": "not_connected"
    }
  ],
  "_ai_hint": "This business uses: Deliveroo, Uber Eats. However, these providers don't have MCP connectors available yet."
}

// AI informs user that ordering via MCP is not yet available

Example 3: Gym / Fitness

// User: "I want to book a class at Basic-Fit"
resolve_domain("basic-fit.com")

// Response
{
  "entity": {
    "name": "Basic-Fit",
    "category": "gym",
    "city": "Amsterdam"
  },
  "connectors": [
    {
      "slug": "basic-fit",
      "name": "Basic-Fit",
      "status": "connected",
      "mcp_endpoint": "https://mcp.basic-fit.com",
      "external_id": "gym-nl-001"
    }
  ]
}

// AI can now help user book a class via Basic-Fit MCP

Adding to AI Assistants

Claude (Desktop & Web)

Both Claude Desktop and Claude.ai use the same remote MCP URL:

https://connect.dockai.co/mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json

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

Claude.ai (Web)

Settings → Connectors
Add custom connector
Paste the URL above

Requires Pro or Max plan

ChatGPT

Go to Settings → Apps
Enable Developer mode in Advanced settings
Click Create under Connectors
Enter URL: https://connect.dockai.co/mcp

Requires Plus, Pro, or Business plan

Mistral Le Chat

Click Intelligence → Connectors
Click "+ Add Connector"
Select "Custom MCP Connector" tab
Enter URL: https://connect.dockai.co/mcp

Available on all plans

Troubleshooting

Domain not found

Verify the domain is correct (without https:// or trailing paths)
The business may not be in our database yet
Try the main domain (e.g., "brand.com" instead of "shop.brand.com")

Connector not working in Claude/ChatGPT/Mistral

Ensure you have a compatible plan (Claude Pro/Max, ChatGPT Plus/Pro, Mistral any plan)
Restart your AI assistant after adding the connector
Check that the URL is exactly: https://connect.dockai.co/mcp

For AI Agents

Quick Start

How it works

Authentication

How OAuth works with MCP

For direct API access

Tool: resolve_domain

Response

Response Fields

Tool: execute_action

Response

Usage Examples

Example 1: Restaurant Booking via Capabilities

Example 2: No MCP Available

Example 3: Gym / Fitness

Adding to AI Assistants

Claude (Desktop & Web)

ChatGPT

Mistral Le Chat

Troubleshooting

Domain not found

Connector not working in Claude/ChatGPT/Mistral

Need help?

Links