MCP Server

MCP server for AI trading agents — discipline as a callable tool

Model Context Protocol server that gives Claude, GPT, and custom LLMs direct tool access to trading discipline evaluation, session health checks, and behavioral detection.

Get MCP access See the engine demo

Best fit

Built for a specific trading workflow

Best for teams building AI/LLM-powered trading agents using Claude Desktop, Claude Code, or custom MCP frameworks.

Why this page exists

AI trading agents need guardrails they can call, not guardrails they need to be told about. The SEIGYO MCP server exposes discipline as four callable tools that any MCP-compatible agent can use before, during, and after trade execution.

What you get

The useful pieces first

Each item is tied to an action the user can understand before they sign up.

evaluate_trade — check a proposed trade against discipline rules before execution

check_session — get current session health, score, budget, and enforcement state

get_rule_types and interpret_verdict — discovery and explanation tools for agent reasoning

Why it matters

The problem this page is solving

This keeps the page focused on trader outcomes instead of a generic product tour.

MCP is the emerging standard for AI agents to call external tools. SEIGYO is the first discipline engine built for it.

The agent gets human-readable explanations of why trades are blocked — improving its decision quality over time.

Runs locally via stdio with no additional infrastructure. Add it to claude_desktop_config.json and the agent has discipline.

Next step

Start with the safest path

The cleanest activation path is still the same: paper broker if you want live guardrails, CSV import if you want historical review, and the demo if you want to understand the system first.

Get MCP access See the engine demo