Model Context Protocol (MCP)

The Proxy MCP server enables AI assistants like Claude, Cursor, and other MCP-compatible tools to interact with your Proxy account directly through natural conversation.

What is MCP?

The Model Context Protocol is an open standard developed by Anthropic that allows AI models to securely connect to external data sources and tools. Instead of copying API documentation and writing code, you can simply ask your AI assistant to perform actions.

Why Use MCP with Proxy?

Natural Language

Ask “Create a card for my shopping agent” instead of writing API calls

Context-Aware

The AI understands your account state and can chain operations intelligently

Secure

OAuth 2.0 authentication ensures your credentials stay safe

Real-Time

Manage cards, agents, and transactions directly from your IDE or chat interface

Example Interactions

Once configured, you can have conversations like:

You: Create a new agent called "Research Assistant" with a $500 monthly limit

Claude: I've created the agent "Research Assistant" with ID agent_abc123.
        It has a spending limit of $500/month. Would you like me to create
        a card for this agent?

You: Yes, create a virtual card for it

Claude: Done! Card created (ending in 4242). The agent can now use this
        card for purchases. Should I configure any spending policies?

Available Operations

The Proxy MCP server exposes 29 tools across these categories:

Category	Description
Onboarding	Start onboarding, check status
KYC	Get verification links, check KYC status
Agents	Create, list, update, pause, disable agents
Cards	Create, list, lock, unlock, cancel cards
Intents	Declare spending intents (agent-only)
Access	Request card credentials (agent-only)
Transactions	View transaction history
Disputes	File and manage disputes
Funding	Get deposit instructions, view funding history
Balance	Check available balance

Authentication Types

The MCP server supports two authentication modes:

User Authentication

Authenticated via Clerk JWT tokens. Users can:

Manage agents and cards
Configure policies
View all transactions
Handle disputes and funding

Agent Authentication

Authenticated via agent tokens (agt_xxx). Agents can:

Create spending intents
Request card access (retrieve credentials)
View their own cards and transactions
Check balance

Architecture

+----------------+     +------------------+     +----------------+
|  Claude/Cursor | --> |  Proxy MCP       | --> |  Proxy API     |
|  (MCP Client)  |     |  Server          |     |  (Convex)      |
+----------------+     +------------------+     +----------------+
        |                      |
        |  JSON-RPC 2.0       |  Authenticated
        |  over HTTP          |  mutations/queries

The MCP server acts as a bridge between AI assistants and the Proxy platform, translating natural language requests into authenticated API calls.

MCP Server

MCP Overview

Model Context Protocol (MCP)

What is MCP?

Why Use MCP with Proxy?

Natural Language

Context-Aware

Secure

Real-Time

Example Interactions

Available Operations

Authentication Types

User Authentication

Agent Authentication

Architecture

Next Steps

Quickstart

Tools Reference

​Model Context Protocol (MCP)

​What is MCP?

​Why Use MCP with Proxy?

Natural Language

Context-Aware

Secure

Real-Time

​Example Interactions

​Available Operations

​Authentication Types

​User Authentication

​Agent Authentication

​Architecture

​Next Steps

Quickstart

Tools Reference

Model Context Protocol (MCP)

What is MCP?

Why Use MCP with Proxy?

Example Interactions

Available Operations

Authentication Types

User Authentication

Agent Authentication

Architecture

Next Steps