The Messages API is Anthropic's primary interface for conversing with Claude, where you send an ordered list of user and assistant messages and receive Claude's reply. It supports system prompts, streaming, tool use, and parameters such as max_tokens and temperature.
The Messages API is Anthropic's primary interface for conversing with Claude, where you send an ordered list of user and assistant messages and receive Claude's reply.
A system prompt is the instruction, passed separately from the conversation, that sets Claude's role, behaviour, and constraints for the whole exchange.
The context window is the maximum amount of text, measured in tokens, that Claude can consider at once — including the system prompt, the conversation, and the response.
A token is the basic unit of text that Claude processes, roughly a few characters or part of a word.
Prompt engineering is the practice of designing inputs — instructions, examples, and structure — to reliably get the output you want from Claude.
Chain of thought is a prompting technique that asks Claude to reason step by step before giving a final answer, which improves accuracy on complex tasks.
Multishot prompting is the technique of including several examples of the desired input-output behaviour in the prompt so Claude generalises the pattern.
Tool use (function calling) is the capability that lets Claude call external functions you define with a name, description, and input schema.
The Model Context Protocol (MCP) is an open protocol that standardises how applications connect Claude to external tools and data sources.
An agent is a system in which Claude directs its own control flow — deciding which tools to call and when — to accomplish a goal, as opposed to a workflow that follows predefined steps.
Prompt caching is a feature that stores a reusable prefix of a prompt so repeated requests reuse it, reducing latency and cost.
A hallucination is a confident but incorrect or fabricated output from a language model.
Prompt injection is an attack in which untrusted input tries to override the developer's instructions to Claude, for example by hiding commands inside a document or webpage.