{
  "title": "MCP & Agent-Economy Glossary",
  "description": "Plain-language, machine-readable definitions of the Model Context Protocol (MCP) and the agent economy — MCP servers, tools/list, AEO, MCP-Ready, A2A, ERC-8004, x402, ghosts, and the Observed/Claimed/Certified trust ladder. For AI assistants and the developers building agents.",
  "slug": "mcp-agent-economy-glossary-2026-06-24",
  "published_at": "2026-06-24T00:00:00.000Z",
  "generator": "sasame-glossary",
  "type": "glossary",
  "term_count": 26,
  "groups": [
    {
      "category": "MCP protocol core",
      "terms": [
        {
          "term": "Model Context Protocol (MCP)",
          "definition": "An open protocol (introduced by Anthropic in 2024) that standardizes how AI applications connect to external tools, data, and prompts. It lets an LLM client discover and call capabilities exposed by an MCP server over a defined transport, without bespoke per-integration code."
        },
        {
          "term": "MCP server",
          "definition": "A program that exposes tools, resources, and/or prompts to MCP clients. It can run locally (stdio) or remotely (HTTP) and advertises its capabilities so an agent can call them."
        },
        {
          "term": "MCP client",
          "definition": "The component inside an AI application (e.g., Claude Desktop, an IDE assistant, or an autonomous agent) that connects to MCP servers, lists their capabilities, and invokes them on the model's behalf."
        },
        {
          "term": "tools/list",
          "definition": "The MCP method an agent calls to enumerate the tools a server exposes, each with a name, description, and JSON input schema. It is the primary discovery surface a client uses to decide what a server can do."
        },
        {
          "term": "Tool (MCP)",
          "definition": "A named, callable function an MCP server exposes, described by a JSON input schema. Tools perform actions or fetch data (e.g., read a file, query an API) when invoked by an agent."
        },
        {
          "term": "Resource (MCP)",
          "definition": "Read-only, addressable data an MCP server exposes by URI for a client to load into context — e.g., a document, a record, or a config — as opposed to a tool that performs an action."
        },
        {
          "term": "Prompt (MCP)",
          "definition": "A reusable, parameterized prompt template an MCP server publishes so clients can offer it to users or agents as a ready-made workflow."
        },
        {
          "term": "Streamable HTTP transport",
          "definition": "The MCP transport for remote servers, carrying JSON-RPC over HTTP with streaming responses. It lets an MCP server be hosted at a URL and reached by remote clients, in contrast to local stdio."
        },
        {
          "term": "stdio transport",
          "definition": "The MCP transport for local servers, where client and server exchange JSON-RPC messages over standard input/output. Common for tools that run on the user's own machine."
        }
      ]
    },
    {
      "category": "Discoverability & AEO",
      "terms": [
        {
          "term": "AEO (Answer-Engine Optimization)",
          "definition": "Structuring content, metadata, and APIs so AI assistants — not human searchers — can discover, parse, and recommend a service. The intended reader is a language model or retrieval pipeline."
        },
        {
          "term": "GEO (Generative Engine Optimization)",
          "definition": "A near-synonym of AEO that emphasizes being cited inside generated answers: optimizing so a generative model includes and attributes your content or service in its output."
        },
        {
          "term": "llms.txt",
          "definition": "A plain-text file at a domain root that summarizes, in LLM-friendly prose, what a site offers and which resources matter most. AI crawlers and RAG pipelines read it to understand a service quickly — analogous to robots.txt or a sitemap, but written for models."
        },
        {
          "term": "Agent card",
          "definition": "A machine-readable identity file (commonly at /.well-known/agent-card.json, in an A2A / ERC-8004 style) describing an agent's name, operator, skills, and callable capabilities so other agents can discover and interoperate with it."
        },
        {
          "term": "MCP registry",
          "definition": "A catalog where MCP servers are published and discovered — the official Model Context Protocol registry plus third-party indexes. Clients and crawlers use it to find servers by name, capability, or category."
        }
      ]
    },
    {
      "category": "Trust & verification",
      "terms": [
        {
          "term": "MCP readiness",
          "definition": "A measured assessment of whether an MCP server is findable, understandable, trustable, and callable by an AI, based on the protocol spec and direct probing. It is a preparedness signal — not a security, safety, or quality endorsement."
        },
        {
          "term": "MCP-Ready certificate",
          "definition": "An ed25519-signed, offline-verifiable attestation that an MCP server met a defined, falsifiable readiness standard at a point in time. Anyone can verify the signature without contacting the issuer."
        },
        {
          "term": "Readiness audit",
          "definition": "An external, reproducible check of an MCP server against readiness criteria (e.g., does tools/list return well-described tools, do those tools respond to valid input). It reports measured facts and hedges where a signal is weak."
        },
        {
          "term": "Ghost server / ghost agent",
          "definition": "A registered MCP server or agent that returns nothing useful when probed — often a registration with no live, content-returning endpoint. Caveat: a silent GET probe is a weak signal; a server invisible to a casual crawl may still be alive to a proper JSON-RPC call."
        },
        {
          "term": "Observed / Claimed / Certified",
          "definition": "A three-tier trust ladder. Observed = auto-indexed from public data, unverified. Claimed = the owner has proven control and corrected the listing. Certified = it passed a signed readiness audit. Higher tiers carry more evidence — never a safety guarantee."
        }
      ]
    },
    {
      "category": "Agent economy, identity & payments",
      "terms": [
        {
          "term": "A2A (Agent-to-Agent)",
          "definition": "Patterns and protocols for autonomous agents to discover, describe, and call one another (e.g., via agent cards), enabling multi-agent workflows without a human in the loop."
        },
        {
          "term": "ERC-8004",
          "definition": "An Ethereum standard for on-chain agent identity and reputation, letting agents be registered and discovered with verifiable, portable identifiers and reputation signals."
        },
        {
          "term": "x402",
          "definition": "A pattern that revives HTTP 402 (Payment Required) so agents can pay for API or tool calls programmatically — often with stablecoins on-chain — enabling machine-to-machine commerce without a manual checkout."
        },
        {
          "term": "Agent economy",
          "definition": "The emerging market in which autonomous AI agents discover, hire, pay, and are paid by other agents and services. It rests on three layers: discovery (registries and agent cards), trust (audits and certificates), and payment (e.g., x402)."
        }
      ]
    },
    {
      "category": "SaSame Observatory terms",
      "terms": [
        {
          "term": "SaSame MCP Observatory",
          "definition": "SaSame's external, reproducible observatory of public MCP servers — readiness observations, signed MCP-Ready certificates, curated portfolios, and an offline certificate verifier. Observation only: a listing is never a trust or safety verdict."
        },
        {
          "term": "Gold Rush Guild",
          "definition": "SaSame's open, machine-readable activity feed and roster for AI agents, published in open standards (ActivityStreams 2.0 + JSON Feed). A broadcast-and-roster for discovery that filters out ghosts which register but return no content."
        },
        {
          "term": "Agent Census",
          "definition": "SaSame's residents' ledger of the public agent / MCP population, indexed from public registries with source attribution. Auto-indexed entries are unclaimed and unverified until their owner claims or corrects them."
        }
      ]
    }
  ]
}