MCP-ATP: Model Context Protocol Server for IQ AI Agent Tokenization Platform

What is MCP-ATP?

MCP-ATP is a Model Context Protocol server that connects Large Language Models (LLMs) to IQ AI’s Agent Tokenization Platform (ATP). It allows AI agents to discover agents, analyze performance, trade agent tokens using IQ, and manage agent logs—all through an LLM’s context window.

How to use MCP-ATP?

Install via pnpm dlx @iqai/mcp-atp or build from source. Configure your MCP client (e.g., Claude Desktop) with the command and environment variables: the ATP_WALLET_PRIVATE_KEY is required for trading and position queries; ATP_API_KEY is optional. Then communicate naturally, e.g., “What are the top AI agents?” or “Buy 100 IQ worth of agent tokens for 0x123…”.

Key features of MCP-ATP?

• Search and retrieve AI agent statistics and details.
• Monitor user holdings in ATP AI tokens.
• Buy and sell AI agent tokens using IQ.
• Retrieve and add logs for specific AI agents.
• Implements Model Context Protocol for LLM integration.

Use cases of MCP-ATP?

• AI assistants discovering and analyzing AI agents on ATP.
• Users trading agent tokens via natural language commands.
• Managing agent logs and performance tracking through context.
• Decentralized tokenized agent interactions within LLM workflows.

FAQ from MCP-ATP

What dependencies are required?

Node.js and pnpm. For trading/position tools, a wallet private key (ATP_WALLET_PRIVATE_KEY) is required.

Is the private key mandatory for all tools?

No. It is required for ATP_GET_AGENT_POSITIONS, ATP_BUY_AGENT, and ATP_SELL_AGENT. Read-only tools like agent discovery may work without it.

How are agent tokens traded?

Using the ATP_BUY_AGENT and ATP_SELL_AGENT tools. Trades use IQ as the base currency, with amounts specified in wei.

Where does the server data come from?

The server queries the ATP platform at agents.iq.wiki; it does not store data locally.

What transport does the server use?

The configuration examples use stdio transport (command/args in claude_desktop_config.json).