mcp-core

# 🧬 MCP Core MCP Core is the reference implementation of the Model Context Protocol — a lightweight, zk- and chain-friendly system for verifiable, composable model context commitments onchain. ### It glues together three simple artifacts: - 🧠 Model Hash — SHA-256 / Blake3 digest of model weights or prompt - 📄 Context Manifest — JSON metadata (datasets, hyperparameters, license, etc.) - 🔐 Proof Stub — zk-friendly cryptographic commitment (Poseidon, Keccak, Rescue) Built for composability, speed, and low-gas operations across EVM and non-EVM chains. ## 🌐 What is MCP? The Model Context Protocol (MCP) defines a minimal, verifiable, and gas-efficient standard for binding AI model metadata to onchain identities. ### It enables features like: - Onchain agent memory - Auditable model registries - Context-linked NFTs - DAO explainers & context badges ## 🔧 Contracts Overview It allows users to: - putContext(bytes32 modelHash, bytes32 contextCID) - getContext(address agent) → (modelHash, contextCID) The contract is chain-agnostic but optimized for low gas on EVM via hash precompiles. ## ✍️ Example: putContext() ``` const modelHash = ethers.utils.sha256("gpt-weights-v1"); const contextCID = ethers.utils.sha256("manifest.json"); await contextStore.putContext(modelHash, contextCID); ``` All context is stored onchain as hash commitments — raw data remains offchain and private. ## 🧠 MCP Use Cases Explore dApps built with MCP: - Contextual GPT Agent w/ Wallet Memory - DAO Proposal Explainer - AI Chat History NFT - Mintable AI Agents - Verifiable Context Badges See roadmap → https://z1labs.ai/roadmap