Checkout Agent

🚧 Coming Soon

Overview

Experimental Notice

The Checkout Agent is currently in active development and should be considered experimental. It is not yet intended for production use. You may encounter errors or unexpected results.

The Checkout Agent is an intelligent automation layer built on top of the MCP (Multi-Context Protocol) server that enables AI-driven checkout automation across e-commerce platforms.

It orchestrates a three-step checkout process — initiate, continue, and complete — allowing autonomous agents to navigate real online stores, add products to carts, fill in shipping and payment details, and finalize purchases.

Designed to function as an upstream MCP service, enabling other MCP servers or autonomous agents to delegate checkout workflows to it, while remaining compatible with any MCP-supported client that uses HTTP Streamable connections.

Why Use It?

Consistency & Reliability: Standardizes checkout automation across multiple sites.
Observability: Includes logging, screenshots, and extracted summaries for full transparency at every stage.
Extendability: Built on MCP, making it easy to integrate or extend with other AI tools or other agents.
Automation: Can operate with minimal human input, guided by intelligent actions and contextual reasoning.

Who It’s For

Merchants and Retailers who want to make their sites MCP-compatible for AI-driven shopping and fulfillment.
Commerce Platforms that wish to provide AI checkout capabilities to their merchants.

Getting Started

The Checkout Agent MCP server can be used with any MCP-compatible client that supports HTTP Streamable such as Claude Desktop, Cursor, or VSCode. This section walks you through connecting to the hosted Checkout Agent.

Prerequisites

Before you begin, ensure you have the following installed:

Node.js (for npx)
Your client must support Streamable HTTP connections.

Quick Start

You can connect to the hosted Checkout Agent MCP server using the mcp-remote package.

The Checkout Agent MCP server is available at:

https://ai-fulfilment-mcp.loginid.io/mcp

You’ll need this URL when configuring your client.

1. Configure your client

Add the following configuration block to your AI client’s MCP configuration file (or MCP settings area):

{
  "mcpServers": {
    "loginid-checkout-agent-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://ai-fulfilment-mcp.loginid.io/mcp"
      ]
    }
  }
}

info

Some clients (like Claude Desktop or Cursor) handle configuration differently — check their documentation for where to paste your MCP server definitions.

2. Restart your client

Once the configuration is added, restart your AI client. You should see a new available MCP server listed as:

loginid-checkout-agent-mcp

Tools Reference

The Checkout Agent exposes the following checkout automation tools via the MCP server.
These tools form a three-stage e-commerce checkout pipeline: initiate, continue, and complete.

initiate_checkout

Initiates the checkout process for a given product by navigating to the product page, adding the item to the cart, and preparing a cart summary.

Parameters:

product_url (required)
- URL of the product page.
- Must be a valid URL (e.g., https://example.com/product/123).

Returns:

status (string)
- Status of the tool execution.
  Possible values: "success", "error".
message (string)
- Human-readable message describing the outcome.
timestamp (string)
- ISO 8601 timestamp of the operation.
request_id (string)
- Unique identifier for this operation.
data (object)
- Contains the output of the initiation step:
  - screenshot (string): URL to the captured screenshot.
  - preview_data_uri (string): Base64-encoded preview image.
  - summary (string): JSON string summarizing order totals.

Example Summary:

{"shipping":"Free","tax":"$0.00","total":"$120.00"}

Example Response Structure:

{
  "status": "success",
  "message": "Checkout initiated successfully",
  "timestamp": "2025-10-28T18:17:26.127Z",
  "request_id": "e8703754-fff7-40e6-a649-d185f8d0eb65",
  "data": {
    "screenshot": "/screenshots/screenshot-BaseAgent-checkout-2025-10-28T18-17-26-077Z.png",
    "preview_data_uri": "data:image/png;base64,...",
    "summary": "{\"total\":\"$120.00\"}"
  }
}

continue_checkout

Continues an existing checkout by navigating to the cart or checkout page and automatically filling shipping and payment details.

Parameters:

cart_url (optional)
- URL of the cart page.
checkout_url (optional)
- URL of the checkout page.
(payment details are managed internally by the agent)

Returns:

status (string)
- Status of the operation ("success" or "error").
message (string)
- Descriptive success or error message.
timestamp (string)
- ISO 8601 timestamp when checkout continuation finished.
request_id (string)
- Unique identifier for this tool execution.
data (object)
- Contains extracted checkout details:
  - screenshot (string): URL of the current checkout screenshot.
  - preview_data_uri (string): Base64-encoded thumbnail.
  - summary (string): JSON string including masked payment data and totals.

Example Summary:

{
  "shipping":"2803 TORO DR SAN MATEO, CA 94403",
  "payment":"************1111",
  "shippingCost":"$19.99",
  "taxCost":"$1.92",
  "totalPrice":"$19.99"
}

Example Response Structure:

{
  "status": "success",
  "message": "Checkout Forms Filled successfully",
  "timestamp": "2025-10-28T18:18:34.465Z",
  "request_id": "e8703754-fff7-40e6-a649-d185f8d0eb65",
  "data": {
    "screenshot": "/screenshots/screenshot-BaseAgent-checkout-2025-10-28T18-18-34-188Z.png",
    "preview_data_uri": "data:image/png;base64,...",
    "summary": "{\"payment\":\"************1111\",\"taxCost\":\"$0.00\",\"totalPrice\":\"$120.00\"}"
  }
}

complete_checkout

Completes the checkout by performing the final action (e.g., "Place Order" or "Pay Now") and verifying that the order was successfully placed.

Parameters:
(none) — must be called after a successful continue_checkout.

Returns Fields:

status (string)
- Operation result status.
message (string)
- Completion message.
timestamp (string)
- ISO 8601 timestamp of the order submission.
request_id (string)
- Unique request identifier.
data (object)
- Contains confirmation information:
  - screenshot (string): URL to the post-checkout screenshot.
  - preview_data_uri (string): Base64-encoded preview.
  - summary (string): JSON string with order confirmation fields.

Example Summary:

{"orderId":"#12345","deliveryDate":"Nov 3, 2025","totalPrice":"$120.00"}

Example Response Structure:

{
  "status": "success",
  "message": "Checkout completed successfully",
  "timestamp": "2025-10-28T18:19:05.842Z",
  "request_id": "e8703754-fff7-40e6-a649-d185f8d0eb65",
  "data": {
    "screenshot": "/screenshots/screenshot-BaseAgent-checkout-2025-10-28T18-19-05-812Z.png",
    "preview_data_uri": "data:image/png;base64,...",
    "summary": "{\"orderId\":\"#12345\",\"deliveryDate\":\"Nov 3, 2025\",\"totalPrice\":\"$120.00\"}"
  }
}

Examples & Use Cases

Initiating Checkout for a Product

Example user prompt:

Add this product to my cart: https://store.example.com/products/sneaker-123

This prompt would trigger the AI assistant to call the initiate_checkout tool.

It will navigate to the product page, add the product to the cart, and extract order summary details such as shipping, tax, and total price.

Triggered Tool: initiate_checkout

Continuing an Existing Checkout

Example user prompt:

Continue the checkout and fill in my shipping and payment details.

This would invoke the continue_checkout tool.

The agent navigates to the cart or checkout page, fills in checkout forms, and provides a summary of the shipping, payment, and totals.

Triggered Tool: continue_checkout

tip

For greater control, users or agents can also specify a known cart or checkout URL using the cart_url or checkout_url parameters.

Completing a Checkout

Example user prompt:

Place the order now.

This prompt triggers the complete_checkout tool.

The agent identifies the primary purchase confirmation button (e.g., “Place Order” or “Pay Now”), submits the order, and verifies that the transaction completed successfully.

Triggered Tool: complete_checkout

Performing the Full Checkout in a Single Step

Example user prompt:

Buy this product: https://store.example.com/products/sneaker-123

In this scenario, the agent may perform the entire checkout process autonomously, calling the following tools in sequence:

initiate_checkout — Adds the product to the cart.
continue_checkout — Fills in checkout and payment details.
complete_checkout — Finalizes and confirms the order.

Recommendation

While the full flow can be executed in one step, it’s often better to call each stage individually. This allows users and other agents to inspect screenshots, summaries, and logs at every phase for better observability and debugging.

Overview​

Why Use It?​

Who It’s For​

Getting Started​

Prerequisites​

Quick Start​

1. Configure your client​

2. Restart your client​

Tools Reference​

initiate_checkout​

continue_checkout​

complete_checkout​

Examples & Use Cases​

Initiating Checkout for a Product​

Continuing an Existing Checkout​

Completing a Checkout​

Performing the Full Checkout in a Single Step​

Overview

Why Use It?

Who It’s For

Getting Started

Prerequisites

Quick Start

1. Configure your client

2. Restart your client

Tools Reference

initiate_checkout

continue_checkout

complete_checkout

Examples & Use Cases

Initiating Checkout for a Product

Continuing an Existing Checkout

Completing a Checkout

Performing the Full Checkout in a Single Step