Claude Overview

What is Claude?

Claude is Anthropic's family of large language models, built with a focus on being helpful, harmless, and honest. It's designed for safety-first usage: long-context reasoning, instruction following, and reliable tool use.

Claude is delivered through several surfaces:

claude.ai — consumer/team chat UI (conversations, Projects, file uploads, Artifacts).
Claude Code — CLI/IDE agent for software engineering.
Claude API — programmatic access via the Anthropic Messages API.
Cloud marketplaces — Amazon Bedrock, Google Vertex AI.
Agent SDK — building custom agents on top of Claude.
Model Context Protocol (MCP) — open protocol Claude uses to talk to external tools/data.

Pillars to remember for cert prep

Pillar	Why it matters
Long context	Claude supports very long contexts (200K+ tokens), enabling RAG over large corpora and long agentic conversations.
Tool use	Claude can decide when and how to call tools, returning structured `tool_use` blocks.
Extended thinking	"Thinking" mode lets Claude reason through hard problems before responding.
Vision	Image input is supported on the multimodal models.
Prompt caching	Marked prefixes are cached server-side for big cost/latency wins on repeated prompts.
Constitutional AI	Anthropic's safety methodology — models trained against a written constitution of principles.

How Claude is built (high level)

Trained with RLHF + Constitutional AI (RLAIF, self-critique against a constitution).
Released in tiered model families: Opus (most capable), Sonnet (balanced), Haiku (fastest/cheapest).
Each generation (3, 3.5, 4, 4.5, 4.6, 4.7) raises capability and lowers cost for equivalent tier.

Claude Overview

What is Claude?

Pillars to remember for cert prep

How Claude is built (high level)

See also