Integration Overview

Architecture

Dock AI acts as a bridge between your existing API and AI assistants. We don't require you to change your API – we build an adapter that translates between your system and the MCP protocol.

Your API

REST, GraphQL, SOAP, or any HTTP-based API

Dock AI Adapter

Translates your API to MCP tools

AI Assistants

Claude, ChatGPT, Gemini, etc.

What We Need From You

To create an adapter for your platform, we need:

• API Documentation – OpenAPI/Swagger spec or detailed endpoint docs
• API Credentials – Service account or API key for server-to-server calls
• Business Logic – Any rules about how your platform works (booking windows, cancellation policies, etc.)
• Test Environment – Sandbox or staging API for testing

Adapter Development

Each partner gets a custom adapter that connects their API to Dock AI. The adapter translates between your specific API and our unified MCP tools (search_businesses, get_availability, book, etc.).

• Maps your API endpoints to unified MCP tools
• Handles authentication and rate limiting
• Validates and transforms data formats
• Provides helpful error messages for AI assistants

// Each partner gets a custom adapter
export class OpenTableAdapter extends BaseAdapter {
  name = 'opentable'
  category = 'restaurant'

  // Maps to search_businesses tool
  async search(query: SearchQuery) {
    const results = await this.api.get('/restaurants', {
      q: query.query,
      city: query.location,
    })
    return results.map(formatBusiness)
  }

  // Maps to get_availability tool
  async getAvailability(businessId: string, date: string) {
    return await this.api.get(`/restaurants/${businessId}/slots`, { date })
  }

  // Maps to book tool
  async book(params: BookingParams) {
    return await this.api.post(`/restaurants/${params.businessId}/bookings`, {
      date: params.date,
      time: params.time,
      covers: params.partySize,
      customer: { name: params.customerName, email: params.customerEmail }
    })
  }
}

User Authentication Flow

When a user wants to perform actions on your platform through an AI assistant, they authenticate via OAuth 2.0:

1. User asks AI to book a table
2. AI requests action via Dock AI MCP
3. Dock AI checks if user is authenticated
4. If not, user is redirected to authorize
5. User grants permissions
6. Action is performed on your platform