Glossary & Terminology

🔧 Core Protocol Terms

Term	Definition
MCP	Model Context Protocol – Standard interface and schema for representing, verifying, and injecting contextual information into LLMs and intelligent agents.
MCP3	Third-generation protocol unifying context handling with Web3 primitives, enabling LLM/agent consumption, zk-proofs, and privacy-aware context orchestration.
MCO	Machine Context Object – Signed, structured, and optionally encrypted payload representing verifiable context for a specific subject (user or agent).
Context	Collection of user or agent-related attributes (identity, activity, state, access rights) that influence model behavior.
Scope	Permission label specifying which part of the context can be disclosed or used (e.g., read.profile, read.dao).
Subject	The identity (typically a DID or Ethereum address) for whom the MCO is created.
Prompt Augmentation	Process of inserting structured or embedded context into a model prompt to ground the LLM's behavior and reasoning.

---

🛡️ Security & Identity

Term	Definition
EIP-712	Ethereum standard for signing typed structured data. Used to authenticate MCOs.
JWS	JSON Web Signature – Alternative signature format for backend implementations.
DID	Decentralized Identifier – Global standard for self-sovereign identity.
ZK Proof	Zero-Knowledge Proof – Prove something is true without revealing the underlying data.
Semaphore	ZK protocol for anonymous signaling. Used in proof-of-humanity and reputation-preserving settings.
RLN	Rate Limiting Nullifier – ZK-based rate limiting tool to prevent abuse in anonymous systems.

---

🔄 Data & Execution

Term	Definition
Timestamp	Unix time indicating when the MCO was created. Used for expiration and freshness checks.
Expires_in	Optional TTL (time-to-live) in seconds, after which an MCO is invalid.
Signer	Wallet or key that signs a context object. Can be user or delegated bot.
Delegated Signer	Scoped, revocable signing key issued by a user to another entity (e.g., API key).
Merkle Tree	Tree of hashes for efficient and secure verification of data integrity. Used for batched attestations.

---

🤖 LLM Integration

Term	Definition
Agent	AI component (usually LLM-based) using contextual information to take actions.
System Prompt	Prompt provided to the LLM to define persona, goals, or constraints. Often where context is injected.
Function Calling	Allows LLMs to call external tools/functions, enhanced by MCP3 context injection.
Context Injection	Embedding signed/structured context into an LLM prompt.

---

📦 Storage & Infrastructure

Term	Definition
IPFS	InterPlanetary File System – Decentralized storage layer for context blobs.
Arweave	Permanent decentralized storage system, optionally used for MCO persistence.
EAS	Ethereum Attestation Service – Smart contract framework for storing attestations on-chain.
Gateway	Server that prepares or verifies MCOs. Can be run by user, DAO, or protocol provider.
Registry	Smart contract or database tracking allowed scopes, signers, or delegations.

---

🧰 SDK Specific Terms

Term	Definition
buildMCO()	Generate and sign a Machine Context Object.
verifyMCO()	Verify a received MCO’s signature, timestamp, and scope.
generatePrompt()	Construct an LLM-friendly prompt using the context.
uploadToIPFS()	Store MCO payloads or claims on decentralized storage.
createDelegatedSigner()	Create a signer key with limited permissions and expiry.
generateZKClaim()	Generate a zero-knowledge proof for an attribute or credential.

---

🧠 Human-AI Interaction

Term	Definition
Contextual Reasoning	LLM’s ability to adapt output based on personalized, verifiable user context.
Grounding	AI model's behavior is "tied" to real-world or structured data (e.g. on-chain facts).
Context Overloading	Risk of inserting too much data into a prompt, leading to token limit overflows or hallucinations.
Context Minimization	Only injecting the most relevant and scoped context for task efficiency and privacy.