Extend AI Toolkit

You can run and connect to an MCP (Model Context Protocol) server that extends Extend AI Toolkit capabilities. This server lets you delegate actions to MCP-compatible clients like Claude Desktop, enabling seamless receipt handling, expense management, and transaction lookups through Extend APIs.

How to use

Set up and run the MCP servers locally to enable AI-driven interactions with Extend’s spend management features. You can operate an MCP server directly from your development environment and, if needed, connect additional MCP components like a filesystem-based server to handle files such as receipts. Use these servers to drive actions like uploading receipts, automatching them to transactions, and querying virtual cards, credit cards, and expense data through Extend APIs.

To integrate with Claude Desktop or other MCP clients, run the provided stdio MCP servers and ensure your environment variables include your Extend API credentials. You can then point your MCP client to the local server instance to start processing user requests via the Extend toolkit APIs.

How to install

Prerequisites before you begin: install Python 3.10 or higher and ensure you have access to an Extend API key and secret.

1. Install the Extend AI Toolkit MCP server dependencies. 2) Set up the MCP server configuration by including the stdio endpoints shown below. 3) Run each server in separate terminal sessions to keep them active.

Install Python packages used by the MCP server (example commands you would run):

pip install extend_ai_toolkit

Configuration and example MCP servers

The following MCP server configurations demonstrate how to wire up the Extend AI Toolkit MCP server and an optional filesystem-based MCP server for receipts. These configurations run locally and expect environment variables for authentication.

[
  {
    "type": "stdio",
    "name": "extend_mcp",
    "command": "python",
    "args": ["-m", "extend_ai_toolkit.modelcontextprotocol.main", "--tools=all"],
    "env": [
      {"name": "EXTEND_API_KEY", "value": "YOUR_API_KEY"},
      {"name": "EXTEND_API_SECRET", "value": "YOUR_API_SECRET"}
    ]
  },
  {
    "type": "stdio",
    "name": "filesystem",
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/receipts/folder"]
  }
]

"language":"json"}]} ,{

## Additional notes

Environment variables shown in the configuration should be replaced with your actual credentials in a secure manner. If you are using a different MCP client, ensure the IPC mechanism and transport align with your client’s capabilities. The Filesystem MCP Server is optional but recommended if you plan to upload receipts and attach them to transactions via Claude Desktop or similar tools.

## Available tools

### get\_virtual\_cards

Fetch virtual cards with optional filters.

### get\_virtual\_card\_detail

Get detailed information about a specific virtual card.

### get\_credit\_cards

List all credit cards.

### get\_credit\_card\_detail

Get detailed information about a specific credit card.

### get\_transactions

Fetch transactions with various filters.

### get\_transaction\_detail

Get detailed information about a specific transaction.

### update\_transaction\_expense\_data

Update expense-related data for a transaction.

### get\_expense\_categories

List all expense categories.

### get\_expense\_category

Get details of a specific expense category.

### get\_expense\_category\_labels

Get labels for an expense category.

### create\_expense\_category

Create a new expense category.

### create\_expense\_category\_label

Add a label to an expense category.

### update\_expense\_category

Modify an existing expense category.

### create\_receipt\_attachment

Upload a receipt and optionally attach it to a transaction.

### automatch\_receipts

Initiate an async job to automatch uploaded receipts to transactions.

### get\_automatch\_status

Get the status of an automatch job.

### send\_receipt\_reminder

Send a reminder by email for a transaction missing a receipt.