Mpc Mcp Server

What is Mpc Mcp Server?

Mpc Mcp Server implements the Metastable Propositional Calculus (MPC) engine, which analyzes the thermodynamic and physical feasibility of logical assertions across multi-step reasoning. It detects epistemic drift and structural conflicts (k-states) by calculating energetic holding costs of maintaining premises over time, used to verify whether a complex sequence of claims can be maintained without contradiction.

How to use Mpc Mcp Server?

Install with pip install -e . (Python ≥ 3.11 required), set environment variables for your chosen AI provider(s) (e.g., ANTHROPIC_API_KEY), then run mpc-server. This starts an MCP server on stdio (registerable in Claude Desktop or any MCP client) and a browser UI at http://localhost:7771. Use the provided MCP tools or the UI to perform MPC analyses.

Key features of Mpc Mcp Server

• Five AI provider routing: Anthropic, Google, OpenAI, Kimi, Ollama.
• Dynamic model listing from supported providers with TTL cache.
• Full MPC analysis with thermodynamic free energy and Ising ground state.
• Per-claim phase assignment with rationale.
• Budget estimation via Theorem 6.1 without API call.
• Browser UI with interactive 3D plots and configuration.

Use cases of Mpc Mcp Server

• Verify multi-step reasoning chains for logical consistency.
• Detect conflicting premises in