Authoryze MCP server
{
"mcpServers": {
"authoryze": {
"type": "http",
"url": "https://authoryze.ai/api/mcp"
}
}
}{
"mcpServers": {
"authoryze": {
"type": "http",
"url": "https://authoryze.ai/api/mcp"
}
}
}{
"mcpServers": {
"authoryze": {
"type": "http",
"url": "https://authoryze.ai/api/mcp"
}
}
}{
"servers": {
"authoryze": {
"type": "http",
"url": "https://authoryze.ai/api/mcp"
}
}
}{
"mcpServers": {
"authoryze": {
"type": "http",
"url": "https://authoryze.ai/api/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.1
What it does
Authoryze is a remote MCP server that lets AI agents request purchases through a human-approval workflow and single-use virtual cards. It is aimed at developers who want to give agents a payment method with defined merchant, category, and spending-limit controls, built on Visa Intelligent Commerce and Mastercard Agent Pay for US-issued cards.
- An agent requests a purchase from a specific merchant and amount, triggering an approval flow before any card details are issued.
- After a purchase is approved, the agent retrieves a single-use virtual card number, expiry, and CVV to complete the transaction.
- A developer checks how much an agent has spent against configured daily, weekly, monthly, or total limits using the spending summary tool.
Tools this server exposes
| Tool | Does |
|---|---|
| request_purchase | Submits a purchase request for a specific merchant and amount, which is then automatically approved, sent to the account owner for approval, or denied. |
| check_status | Looks up the current status of a previously submitted purchase request without exposing any card details. |
| get_spending_summary | Returns how much has been spent against configured limits (daily, weekly, monthly, and total) for the agent and account. |
| retrieve_card | Returns the card number, expiry, and security code for an approved purchase; can only be called once per approval. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.