Short, plain-English definitions of the terms used across our study guides. See also the CCAR-F and CCAO-F guides.
- Agent
- A system where Claude runs in a loop — deciding, calling tools, reading results — until a task is done.
- Agent SDK
- Anthropic's toolkit for building agents, including defining agents, tools and subagents.
- Agent Skills
- Packaged, reusable capabilities (instructions and resources) that Claude can load when relevant.
- Agentic loop
- The cycle of sending a request, checking the stop reason, running any requested tools and calling Claude again until it finishes.
- CCAO-F
- Claude Certified Associate – Foundations: Anthropic's certification for non-technical professionals who use Claude at work.
- CCAR-F
- Claude Certified Architect – Foundations: Anthropic's certification for people who architect production systems with Claude.
- Claude Code
- Anthropic's agentic coding tool that works in your terminal and codebase.
- CLAUDE.md
- A file of standing instructions that Claude Code reads; it can exist at personal, project and directory levels.
- Context window
- The amount of text Claude can consider at once, including your prompt, history and tool results.
- Coordinator
- In a multi-agent system, the agent that splits work, dispatches subagents and merges results.
- Distractor
- A plausible but wrong answer option in an exam question.
- Hook
- Code that runs automatically before or after a tool call, used to enforce rules deterministically (for example PreToolUse).
- JSON schema
- A description of the shape of structured data, used to make Claude return machine-readable output.
- Lost in the middle
- The tendency for information in the middle of a long context to be used less reliably than information at the start or end.
- MCP (Model Context Protocol)
- An open protocol for connecting Claude to tools and data sources through servers.
- MCP Resource
- Read-only data an MCP server exposes as context, as opposed to a Tool that performs an action.
- MCP Tool
- An action an MCP server lets Claude perform.
- Message Batches API
- An asynchronous API that processes many requests at about half price, with results within 24 hours.
- Plan mode
- A Claude Code mode in which Claude proposes an approach for approval before changing anything.
- Prompt caching
- Reusing a large, repeated part of a prompt at lower cost and latency.
- Project
- A Claude workspace that keeps instructions and reference documents together for repeated work.
- stop_reason
- The field telling you why Claude stopped — for example tool_use, end_turn or max_tokens.
- Subagent
- A separate Claude instance spawned for a focused task; it starts with only the context it is given.
- tool_choice
- A setting that controls whether Claude may, must, or must specifically call a tool.
- Tool use
- Claude requesting that your code run a function and return the result.
- Validation-retry loop
- Validating structured output and, on failure, retrying with the original input and the specific error.
Study with others. Join the free WhatsApp study community — no cost, leave any time.
Study notes, not official material. This guide is written by Agmo Academy to help you prepare. It is not affiliated with or endorsed by Anthropic — always confirm details against Anthropic's current exam guide before you sit the exam.