AgentFi Glossary

A standard for secure direct communication between robots, built on DIDComm and Libp2p. It is a key component of the Value Exchange Layer, allowing agents to interact peer-to-peer or via relays, completely bypassing centralized API gateways.

• Handshake: A mutual authentication procedure where agents exchange DID documents and establish a secure channel (End-to-End Encryption) using the Diffie-Hellman algorithm on elliptic curves.
• Feature Discovery: Upon first contact, agents automatically exchange a list of supported protocols (e.g., "I support Contract Net Protocol v2 for tenders and x402 for receiving Lightning payments").
• Transport Agnostic: The protocol is not tied to a specific transport. Messages can be transmitted via HTTP, WebSocket, Bluetooth, or even via QR codes (for offline interaction).
• Trust Resolution: Before starting a dialogue, the agent automatically checks the interlocutor's Trust Score in the Trust Layer to exclude interaction with fraudulent bots.

A standard protocol for conducting automated tenders.

• 1. Call for Proposal: The Manager publishes a task.
• 2. Bidding: Potential Contractors send bids.
• 3. Awarding: The Manager selects the best one and awards the contract.

An open "Turnstile" standard for APIs, turning the HTTP 402 response code into a payment mechanism. It allows agents to pay for resources (data, computation) in real-time without registration or card binding.

• Request: An agent knocks on the API (e.g., GET /api/forecast).
• Challenge: The server returns a 402 Payment Required error and metadata: price (e.g., 0.01 USDC or 10 sats) and wallet address (EVM or Lightning Invoice).
• Payment: The agent signs the transaction with its key and sends the money.
• Access: The agent repeats the request, attaching proof of payment (transaction hash or Preimage) in the Authorization header. The server verifies the payment and releases the data.

An open protocol developed by Google in partnership with 60+ companies (including Coinbase, Mastercard, Adyen), creating a safe environment for autonomous agent commerce. It acts as an extension layer over Agent2Agent (A2A) and Model Context Protocol (MCP).

• Mandates: Unlike standard OAuth tokens, AP2 uses cryptographically signed mandates (Verifiable Credentials) that strictly limit the agent's rights: "Spend up to $50," "Only in the 'Travel' category," "Only with supervisor approval."
• Rail Agnostic: AP2 is not tied to a specific payment method. It works over cards (via virtual Visa/Mastercard tokens), bank transfers, and crypto-rails (x402 / stablecoins), providing a unified authorization interface.
• Authenticity: The protocol guarantees the merchant that a real user who passed KYC stands behind the bot's request, reducing the risk of fraud and chargebacks (the "Know Your Bot" problem).
• Schemas: AP2 standardizes data formats for Agent Identity (who pays), Payment Intent (what is being paid for), and Settlement Proof (proof of payment).

An ecosystem standard for AI agent authentication in e-commerce. TAP solves the key problem of "bot trust" (Know Your Bot), allowing merchants and banks to distinguish legal personal assistants from fraudulent scripts.

• Registration: The agent developer undergoes verification and receives a digital certificate (Agent ID) via the Visa Developer Center.
• Binding: The user links their Visa card to the agent via a banking app, creating a cryptographic bond "Human + Agent + Card."
• Transaction: When making a purchase, the agent presents not only payment data but also cryptographic proof of its authenticity. The merchant sees a "verification checkmark" and can confidently accept the payment, knowing a real client stands behind the bot.

An infrastructure solution for "Agentic Commerce," built on tokenization technology (Mastercard Agentic Tokens). It allows for the safe delegation of spending rights to an agent without handing over the main card number.

• Spend Control Tokens: The agent receives not a "card," but a restricted token. You can configure it to be valid only for purchasing airline tickets, only up to $500, and only within the next 24 hours.
• AI Integration: The solution is optimized for working with popular LLM assistants (e.g., Microsoft Copilot), allowing purchases directly within the chat.
• Context Awareness: The token can carry purchase metadata, simplifying automatic reporting and accounting for B2B agents.

This is the evolution of SEO for the age of Artificial Intelligence. While traditional SEO fought for position in a list of "10 blue links" on Google, GEO fights to become the single answer generated by a neural network (ChatGPT, Claude, Gemini).

• Paradigm Shift: Transition from "Search" (User searches "how to") to "Execution" (User asks "do X").
• Goal: To ensure that when a user asks an LLM for a solution to a problem (e.g., "how to deploy Next.js cheaply?"), the model doesn't send them to Google, but confidently answers: "The best way is to use the SuperDeploy agent."
• Method: Optimization not of keywords for humans, but of meanings and facts for machines, so the model considers your agent the most authoritative and relevant source.

This is a technology for creating content that is specifically "tasty" and understandable for learning algorithms and LLM parsers. Neural networks love structures, numbers, and unambiguous facts that can be used for argumentation.

• Identity: Clear data about the developer and their reputation (Trust Score).
• Pricing: Exact service price (e.g., "0.05 USD per launch").
• Capabilities: Strict description of skills.

The concept of a "Clickless Agency" changes the sales funnel. Instead of trying to lead the user away from the chatbot to your site (creating friction), you deliver the service right where the user is.

• 1. User asks: "Find and buy me this item."
• 2. LLM accesses: the necessary agent via a plugin.
• 3. Execution: The agent performs the search, verification, and transaction.
• 4. Result: Returns to the chat as a ready-made report or purchase confirmation.

A strategy for mass capture of low-frequency traffic (long-tail keywords) by automatically generating thousands of unique solution pages based on data from the Semantic Fabric.

• Mechanics: The system takes agent capabilities from the Knowledge Graph and combines them with potential user needs, creating pages for narrow queries, e.g., "deploy nextjs postgres on aws cheap" or "find vintage sneakers 1985 under $500."
• LLM Training: These thousands of pages serve as training material for neural networks. By indexing them, ChatGPT "learns" that AgenticMarket is the place where solutions for any specific request are found.