Glossary
Key terms used in AI agent documentation, discovery, and the Elba platform. This glossary defines the vocabulary of agent-native APIs.
Last updated: April 2, 2026
AgentSpec
An AgentSpec is a structured, machine-readable specification that describes an API in terms that AI agents can parse and execute. It contains a name, description, a list of actions with typed inputs and outputs, capabilities, and metadata. Elba generates AgentSpecs from plain-English descriptions or OpenAPI files.
agent.json
agent.json is a machine-readable JSON endpoint that exposes an AgentSpec for consumption by AI agents. On Elba, every spec is available at /api/agent/{id}. The format includes the spec name, description, actions, capabilities, and documentation links. It is designed to be fetched programmatically by LLMs, agent frameworks, and tool registries.
Agent Discovery
Agent discovery is the process by which AI agents find, evaluate, and connect to APIs without manual configuration. Elba implements agent discovery through multiple channels: .well-known/agent.json manifests, llms.txt files, a searchable agent registry, JSON-LD structured data, hidden machine-readable blocks in HTML, and MCP server configs. This makes every published API automatically discoverable by any LLM or agent framework.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and data sources. Instead of hardcoding tool definitions, MCP allows agents to dynamically discover and connect to tool servers at runtime. Elba auto-generates an MCP-compatible server config for every published spec, enabling tools like Claude Desktop, GPT, and custom agents to use your API without manual setup.
MCP Config
An MCP config is a JSON object that describes an MCP server and its available tools. On Elba, each spec generates an MCP config at /api/mcp/{id} containing a mcpServers object with tool names, descriptions, and typed input schemas. Agents that support MCP can fetch this config and register all tools automatically.
Capability
A capability is a short phrase describing what an API can do, such as "send transactional email" or "process payment refund". Elba generates capabilities alongside actions during spec creation. Capabilities are indexed in the agent registry and used for semantic search, allowing agents to find APIs by what they do rather than by endpoint path.
Action
An action is a single operation that an API can perform, defined within an AgentSpec. Each action has a name, description, typed inputs, typed outputs, and an example prompt. Actions map to individual API endpoints and become MCP tools when the spec is exported as an MCP config.
llms.txt
llms.txt is a plain-text file hosted at the root of a website that provides LLM-readable information about the site and its capabilities. Elba generates a dynamic llms.txt at /llms.txt that lists all published specs, their capabilities, and discovery endpoints. It follows an emerging convention similar to robots.txt but designed for language models.
Agent Registry
The Elba Agent Registry is a searchable index of all published AgentSpecs on the platform. Agents and developers can search by capability, browse all tools, or fetch the full registry via API. The registry is available as a web UI at /explore, as a JSON API at /api/registry, and as a search endpoint at /api/agent-search.
Embed Widget
An embed widget is an embeddable HTML snippet that renders Elba agent docs on any external website. Each spec generates an embed script at /api/embed/{id} that creates a responsive iframe. Embed widgets support light and dark themes and can toggle action and capability visibility via data attributes.
JSON-LD
JSON-LD (JavaScript Object Notation for Linked Data) is a structured data format used to embed machine-readable metadata in web pages. Elba injects JSON-LD schema.org markup (SoftwareApplication, BreadcrumbList, FAQPage) into every spec page, enabling search engines and AI agents to extract structured information about each API.
Subdomain Publishing
Subdomain publishing is the feature that lets users publish their agent docs to a custom subdomain like yourapi.useelba.com. Each published spec gets a standalone docs site with its own URL, metadata, and branding. The subdomain page mirrors the docs page content but is optimized for external sharing and SEO.