Glossary

34 terms and standards that matter for AI-native software buying — MCP, agent-readiness, AEO, x402 agent payments, and more. Each entry is a short, plain-language definition, not a marketing pitch.

• Agent Orchestration Framework

  A library or runtime for composing multiple LLM calls, tools, and sub-agents into a single reliable workflow — handling state, retries, and hand-offs between steps.

• Agent Test Report

  A Crail review type unique among software directories: an AI agent runs a fixed, pre-published task suite against a product and the full transcript is published, rather than a person's written opinion.

• Agent-Readiness Score

  Crail's published, 0-100 composite score measuring how usable a piece of software is for an AI agent to discover, evaluate, and buy — not how good the product itself is.

• AI Coding Agent

  Software that autonomously plans, writes, tests, and often ships code changes from a natural-language instruction, going beyond simple autocomplete.

• AI Gateway

  A proxy layer that gives an application a single API to route requests across many LLM providers, with fallbacks, load balancing, caching, and cost/usage tracking.

• AI Meeting Assistant

  Software that joins a video call to record, transcribe, and summarize it — typically also extracting action items and syncing notes into a CRM or task tool.

• AI SDR (Sales Development Representative)

  Software that automates the top-of-funnel sales work a human SDR does — prospecting, personalized outreach, and follow-up sequencing — rather than assisting a rep who does it manually.

• AI Support Agent

  A customer-facing AI agent that resolves support requests end-to-end — not just suggesting a reply for a human agent to send, but autonomously answering, taking actions (refunds, order lookups), and escalating only what it can't handle.

• AI Writing Assistant

  Software for drafting, editing, or generating marketing and content copy at volume — distinct from a general-purpose chatbot in that it's built around brand voice, templates, and content workflows.

• Answer Engine Optimization (AEO)

  The practice of structuring content so AI answer engines (ChatGPT, Claude, Gemini, Perplexity) can accurately extract and cite it in conversational answers — as distinct from optimizing for traditional search-result rankings.

• Business Intelligence (BI)

  Software for turning raw business data into dashboards, reports, and ad-hoc queries that non-engineers can use to make decisions — the visualization/reporting layer on top of a data warehouse.

• Claude Skill

  A packaged set of instructions (and optionally scripts/resources) that extends what Claude can do for a particular task, invoked by name inside a conversation.

• Computer-Use Agent

  An AI agent that operates a browser or desktop directly — clicking, typing, and reading the screen — rather than calling a dedicated API, so it can complete tasks on sites and apps with no integration built for it.

• CRM (Customer Relationship Management)

  The system of record for a company's contacts, deals, and customer interactions — the shared source of truth sales, marketing, and support teams work from.

• Generative Image Model

  A model that produces new images from a text prompt (or reference image) — used for anything from marketing creative to product mockups to design iteration.

• Helpdesk Software

  The ticketing system a support team uses to receive, route, and resolve customer requests across channels (email, chat, social) from one shared inbox.

• HRIS (Human Resources Information System)

  The system of record for employee data, payroll, and benefits administration — the HR-side equivalent of a CRM for the employee relationship instead of the customer relationship.

• Human-Verified Review

  A Crail review from a real buyer, tiered by how strongly their identity and product usage were verified before publication.

• Identity and Access Management (IAM)

  Software that controls who can log into which systems and what they're allowed to do once inside — authentication (proving who you are) plus authorization (what you can access).

• JSON-LD / Structured Data

  A JSON-formatted way of embedding schema.org markup in a webpage so search engines and AI systems can read specific facts (price, rating, FAQ answers) unambiguously, rather than inferring them from prose.

• LLM Observability

  Tooling that traces, monitors, and evaluates LLM and AI-agent applications in production — capturing prompts, completions, cost, latency, and quality scores per request.

• llms.txt

  A community convention (not a formal web standard) for a plain-Markdown file at a site's /llms.txt root, giving AI agents a curated index of the site's most important pages.

• Marketing Automation Platform

  Software for running multi-step, triggered marketing campaigns (email, SMS, push) based on customer behavior — instead of one-off broadcast sends.

• Model Context Protocol (MCP)

  An open protocol, introduced by Anthropic in November 2024, that lets AI applications connect to external tools, data sources, and services through a standard client-server interface.

• Model Inference Hosting

  A hosted API that runs a machine learning model (open-weights or custom) and returns predictions on demand — the layer between "I have a model" and "my product can call it."

• Observability (DevOps)

  The combination of logs, metrics, and traces that lets an engineering team understand what a running system is actually doing — the infrastructure/application counterpart to LLM Observability.

• Retrieval API

  A hosted API that takes a query and returns relevant, structured web content or search results — built for LLM/agent consumption rather than for rendering a human-facing results page.

• Self-Serve Transactability

  Whether a buyer (human or AI agent) can complete a purchase of a piece of software entirely online, without a human sales call.

• Synthetic Data

  Artificially generated data that mimics the statistical properties of real data, used to train or test models without exposing real (often sensitive) records.

• Text-to-Speech (TTS)

  Models that convert text into natural-sounding spoken audio — the counterpart to speech-to-text transcription — used for voice agents, narration, and voice cloning.

• Text-to-Video Generation

  Models that generate video clips from a text prompt or a reference image — used for ads, social content, and increasingly for avatar-presenter (talking head) video.

• Vector Database

  A database built to store and search embeddings — numerical representations of text, images, or other content — by similarity, rather than exact match. The core infrastructure behind most RAG systems.

• Work Management Software

  The shared system for tracking tasks, projects, and their status across a team — spanning lightweight kanban boards through to full project-portfolio management for larger organizations.

• x402 (HTTP 402 payment protocol)

  A Coinbase-originated protocol built on the long-dormant HTTP 402 "Payment Required" status code, letting a server request a stablecoin micropayment before an API call completes.