Claude — Chat & Setup
Claude is Anthropic’s AI assistant. This is the on-ramp for the Claude track: what Claude is, the surfaces you’ll reach it through, and the handful of setup steps that make it dramatically more useful than a plain chat box.
Claude is a friendly, careful robot you talk to by typing. You can hand it files to read, save a “project” so it remembers what you’re working on, give it a “voice” so it writes the way you like, and even plug it into your other apps so it can help with them. It’s like a very well-read helper who’s great at explaining things and writing.
What Claude is — and the model tiers
☺ Like you’re 10: Claude comes in three sizes, like picking a bike for a trip: the big powerful one for the hardest hills, a comfy everyday one for most rides, and a light zippy one when you just need to go fast and cheap.
Claude is a family of large language models trained by Anthropic with a strong emphasis on being helpful, honest, and careful. In practice you’ll pick between three tiers, trading speed and cost against raw capability:
| Tier | Best for |
|---|---|
| Opus | The most capable — hard reasoning, long agentic tasks, tricky code. |
| Sonnet | The balanced workhorse — fast and strong for most day-to-day work. |
| Haiku | The quickest and cheapest — high-volume, latency-sensitive jobs. |
Version numbers advance quickly (Opus 4.x, Sonnet 4.x, and so on), so treat any specific number as “true at time of writing” and check the current lineup. Plans run from Free and Pro up through higher consumer and organization tiers (Max, Team, Enterprise), which raise usage limits and add admin and data controls.
Four surfaces, one model
☺ Like you’re 10: The same helper shows up in different doorways — like a friend you can text, invite into your workshop, or hand a big list of chores to. It’s one friend, but which door you use changes how much they do on their own.
“Claude” isn’t a single app — it’s a model you reach through several surfaces, each a different point on the autonomy spectrum. Knowing which surface fits the task is half the skill.
| Surface | What it’s for | Covered in |
|---|---|---|
| The chat app (this page) | Conversation, writing, analysis, quick code — with files, projects, and artifacts. | here |
| Claude Code | An agent in your terminal/editor that plans and edits across your codebase. | Claude Code |
| The Claude API | Build your own app on Claude — messages, tools, streaming, agents. | Building with Claude |
| Cowork | Hand off longer multi-step knowledge work for Claude to run for you. | below |
This is the autonomy spectrum again: the chat app keeps you in the loop turn by turn; Claude Code and Cowork hand off more; the API is where you decide exactly how much to delegate. Same idea as Copilot’s completions → chat → agents, just Anthropic’s surfaces.
The chat features that actually matter
☺ Like you’re 10: These extras are like giving your helper a backpack: a folder that remembers your project, a note about how you like things written, a workbench to build stuff on, and a way to peek into your other apps — so you don’t have to explain everything again.
The free-text box is the least interesting part. These are the features that change how much Claude can do for you:
- Projects — a persistent workspace with its own knowledge (docs, instructions) that every chat in it can see. This is where “Claude remembers my context” comes from.
- Styles — teach Claude a writing voice (your tone, formatting rules) and reuse it, instead of re-explaining every time.
- Artifacts — when Claude produces something substantial (code, a document, a small app), it opens in a side panel you can iterate on, run, and download.
- Files & web search — upload PDFs, spreadsheets, and images for Claude to read; let it search the web for current information.
- Connectors — link Claude to external apps and data. Under the hood these are MCP servers, the same open standard you met earlier.
- Memory — across conversations, Claude can carry forward useful context so you’re not starting cold each time.
Foxy: Delphi, why do I keep re-explaining my project every single chat?
Delphi the Dolphin: You don’t have to! Drop your docs into a Project and I’ll read them every time. Ask, and I’ll build the answer right in an Artifact panel.
Ellie the Elephant: That’s the part I love — a Project remembers your context, so nobody starts cold.
Professor Owl: And pick the right size for the job: Opus for the hardest thinking, Sonnet for everyday work, Haiku when you just need it fast.
Setup that pays for itself
Two small habits do most of the work:
Create a Project for something real you’re working on. Drop in the key docs and write a short instruction like “You’re helping me build X; prefer concise answers and cite the docs.” Then start a chat inside it and notice how much less context you have to repeat. That project instruction is Claude’s version of a project-wide convention — and the direct ancestor of CLAUDE.md in the next page.
Cowork — handing off the longer jobs
☺ Like you’re 10: Cowork is like giving a big group-project task to a helper and saying “go do all the steps and show me when you’re done,” instead of asking one tiny question at a time.
When a task is bigger than a single reply — “research these five vendors and draft a comparison,” “clean up this dataset and chart it” — Cowork lets Claude run it as a multi-step job, using tools and files, and come back with a result you review. It’s the knowledge-work counterpart to Copilot’s cloud agent: you delegate, it works, you check.
You can name the three model tiers and when to use each, the four surfaces Claude comes in, and at least three chat features (projects, styles, artifacts, connectors, memory) that raise what Claude can do. Next: Claude Code, where Claude becomes an agent in your codebase — or jump to the CCAR-F certification if you’re aiming at the exam.
Check your answers
- The three model tiers: Opus is the most capable, for hard reasoning, long agentic tasks, and tricky code; Sonnet is the balanced workhorse, fast and strong for most day-to-day work; Haiku is the quickest and cheapest, for high-volume, latency-sensitive jobs.
- The four surfaces: the chat app (conversation, writing, analysis, quick code with files, projects, and artifacts), Claude Code (an agent in your terminal/editor that edits across your codebase), the Claude API (build your own app on Claude), and Cowork (hand off longer multi-step knowledge work). They sit along the autonomy spectrum, from turn-by-turn chat to more delegated work.
- Chat features that raise what Claude can do: Projects give a persistent workspace whose docs and instructions every chat can see; Styles teach and reuse a writing voice; Artifacts open substantial output in a side panel to iterate on, run, and download; Connectors link Claude to external apps and data via MCP; and Memory carries useful context across conversations. Naming any three (also Files & web search) meets the checkpoint.