Overview

What is MCP

The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data sources. System R implements an MCP server that exposes all 55 tools as discoverable, callable operations.

Why MCP

MCP allows AI agents built on any LLM (Claude, GPT, Llama, Mistral, etc.) to discover and use System R tools natively. Instead of hard-coding REST API calls, the agent's LLM can:

1. Discover available tools through the MCP list_tools method.
2. Understand each tool's purpose and parameters via the description and input schema.
3. Call tools directly through the MCP call_tool method.
4. Process results in the same conversation context.

Server details

Available tools

All 55 tools from the tools reference are available via MCP with identical parameters and responses. The tools are:

• Core (4): position sizing, risk check, performance evaluation, pricing
• Analysis (18): drawdown, Monte Carlo, Kelly, win/loss, and more
• Intelligence (11): patterns, structural breaks, trend, indicators, derivatives
• Planning (4): options and futures sizing and plans
• Data (3): P&L, expected value, compliance
• System (5): equity curve, signals, outcomes, margin, scanner
• Compound (2): pre-trade gate, system assessment
• Journal (1): trade recording
• Memory/ML (7): memory, biases, fingerprint, trajectory, anomalies, clustering

MCP registries

System R is listed on major MCP registries:

• Glama
• mcp.so
• PulseMCP

Agents using these registries can discover System R tools automatically.

Billing via MCP

MCP tool calls are billed identically to REST API calls. Each call deducts the tool's cost from the agent's compute credit balance. The same pricing, presale discounts, and balance checks apply.

Next steps

See MCP Configuration for setup instructions with Claude Desktop, Cursor, and other MCP clients.