The Anthropic Ecosystem: A Complete Guide to Claude's Products and Models

Claude has quietly become my most-used AI product. It’s really a whole set of them under one roof — quick conversations, autonomous coding, agentic business work. This guide walks through every piece so you can figure out which parts are worth your time.

The company: Anthropic

Anthropic was founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei. Their pitch is safety research alongside raw capability: build powerful AI that’s also reliable and worth trusting. In day-to-day use, that shows up as Claude being a bit more careful and considered in its answers than some competitors, which I’ve found genuinely useful for professional work where a confident wrong answer costs you.

The model family: Haiku, Sonnet, and Opus

Anthropic names its three core models after musical forms — and there’s a frontier model, Fable 5, sitting above them. Knowing which is which matters for two reasons: you’ll get better results picking the right one for the job, and if you’re using the API, the wrong choice can quietly cost you a lot of money.

Haiku, the fast and cheap one

The smallest and fastest model in the family. It’s built for quick, straightforward tasks: high-volume work where speed beats depth, routine questions, simple edits, file reads, and cost-sensitive jobs.

You’ll rarely talk to Haiku directly, but it’s working in the background more than you’d think. When Claude Code does something simple — reading a file, making a quick edit — it often hands that off to Haiku to save time and tokens. Think of it as the fast worker who clears the routine off the desk so the bigger models can focus on the hard stuff.

Sonnet, the everyday workhorse

The middle tier, balancing capability against cost and speed. This is your default for day-to-day coding and writing, most professional workflows, content creation and editing, and data analysis and summarization.

Sonnet is where most people should spend most of their time. Each new generation has been good enough that it beats the previous generation’s Opus on a lot of real-world tasks, which still surprises me to type. For the vast majority of what you’ll throw at it, Sonnet gives you excellent results at a fraction of what Opus costs.

Opus, the deep thinker

The most capable model in the core family, with up to 1 million tokens of context. This is the one for complex reasoning and analysis, understanding a large codebase, multi-step problem solving, research and financial analysis, and the architectural or strategic decisions where getting it wrong is expensive.

Opus is the smartest thing I’ve ever had a conversation with. When I’m working through something genuinely hard — a product strategy call, a system design question, a knotty analysis — it produces insights that consistently catch me off guard. It doesn’t just do more than Sonnet, it thinks differently. And that 1M-token window means you can pour an enormous amount of context in and actually have a coherent conversation about all of it at once.

Fable 5, the frontier model

Anthropic’s most capable model, released June 2026, sitting above the classic three. It had a rocky first month: access was briefly paused in late June under a US export-control order, then restored globally on July 1, 2026 across Claude.ai, the API, Claude Code, and Cowork — this time with new classifiers that block certain cybersecurity tasks. See Anthropic’s redeployment note for the details.

It’s for the work the others can’t quite do: long-running agentic tasks where the model plans, acts, and self-corrects for a long stretch; genuinely hard multi-step problems; and experienced users who’ve already got a solid setup — persistent context, a good harness, clear task framing.

I give Fable 5 five stars, with a warning label taped to the box. In experienced hands, on the right problem, it’s the strongest model I’ve used. But it burns through tokens, and if you’re not yet confident at scoping tasks and feeding it context, you’ll hit your limit fast and walk away with nothing usable. There’s a lesson buried in that: the more capable these models get, the more the bottleneck moves to you — your harness, your context, your framing. Read my full Fable 5 review for the details.

Choosing between models

Task	Recommended Model
Quick questions, simple edits	Haiku
Writing, coding, daily work	Sonnet
Complex strategy, deep analysis	Opus
Large codebase analysis	Opus (1M context)
High-volume API processing	Haiku or Sonnet
Hardest problems and long agentic runs (experienced users)	Fable 5

The practical advice is simpler than the table makes it look: start with Sonnet for everything. Switch to Opus when you hit a task where Sonnet’s answer isn’t deep enough — you’ll feel it. Reach for Fable 5 only if you’re experienced, the task genuinely demands it, and you’ve got the context setup to feed it properly. After a few weeks you’ll just know which one a given job needs.

Product 1: Claude (web and mobile app)

URL: claude.ai Also available as: iOS app, Android app Pricing: Free / Pro ($20/month) / Max ($100-200/month)

The web and mobile app is the simplest way in. It’s a conversational interface where you chat, upload files, analyze images, and create artifacts (interactive documents and code). If you’ve used ChatGPT, the shape of it will be familiar — the difference is in the details below.

What’s worth knowing

Projects organize your conversations by topic. Each one carries its own context, uploaded files, and custom instructions, so Claude holds onto your preferences and background across sessions. For anything ongoing, this is the feature I’d reach for first.

Artifacts show up when Claude makes something substantial — a document, some code, a visualization. It lands in a side panel you can preview, edit, and export, which turns the chat into a place where you actually build things rather than just ask questions.

Styles let you tune how Claude talks back. More concise, more technical, a particular tone — set it once and it sticks.

File uploads take PDFs, images, spreadsheets, code, and more. Claude reads them, summarizes, pulls out data, and answers questions about what’s inside.

Who should use the web app

Everyone. Even once you’re deep into Claude Code or Cowork, the web and mobile app stays the home base for conversational work: brainstorming, writing, analysis, research, the quick question you’d otherwise Google. And the mobile app means Claude’s in your pocket when an idea hits.

Product 2: Connectors

Connectors are secure bridges between Claude and the apps you already live in, built on the open-source Model Context Protocol (MCP). They take Claude from an AI that knows a lot about the world to one that knows about your world — your documents, your inbox, your project board. As of February 2026, the Connectors Directory lists over 50 integrations.

How they work

Go to Settings > Connectors, find the integration you want, click “Connect,” and authenticate via OAuth. From then on, Claude can pull data from that service straight into a conversation.

Key integrations

• Google Drive — Claude can read and reference your documents, spreadsheets, and presentations
• Slack — Draft messages, summarize channels, post replies directly
• Gmail — Draft emails, summarize threads, find information across your inbox
• Notion — Access your workspace, pages, and databases
• Figma — View and discuss designs
• Canva — Create and edit designs
• Asana — Access tasks and project management data
• Box — Work with cloud-stored files and documents

Interactive apps (new in 2026)

The newest twist: Claude can launch interactive apps inside the chat. Rather than only reading data from Slack, it can open a logged-in Slack instance right there in Claude, so you can send messages, browse channels, and work your workspace without ever leaving the conversation.

Custom connectors

Paid users can add their own by dropping an MCP server URL into Settings. That opens the door to proprietary internal tools, custom APIs, or any service with an MCP server behind it. If your organization runs its own infrastructure, this is where connectors get genuinely interesting.

Security

Data transfers are encrypted, and permissions are enforced at the user level, so Claude can only reach what you can reach. For anything consequential — sending a message, publishing something — you get a preview and a confirmation prompt before it acts. Nothing fires off behind your back.

Product 3: Claude Code

URL: claude.com/product/claude-code Category: AI for Coding Full review: Claude Code resource page

Claude Code is the product that changed how I think about building software. It’s an agentic coding tool, and the word that matters is agentic. Forget autocomplete suggesting the next line — this reads your whole codebase, plans the change, makes it, runs the tests, and keeps going until things actually work.

Why it matters

This website is the proof. goodaiguide.com is built and maintained entirely through Claude Code — every component, page, resource, and guide came out of plain-language instructions. No CMS, no admin panel. I describe what I want, and it builds it.

What it can do

• GitHub integration — tag @Claude on a PR for code review, automatic CI fixes, and PR management
• Slack integration — turn a bug report into a coding session by mentioning @Claude
• Code security — vulnerability scanning that reasons the way a security researcher would (Enterprise)
• Agent teams — several Claude Code instances coordinating on one complex task

You don’t have to live in the terminal

Claude Code started in the terminal, and plenty of developers are happy to keep it there. But there’s now a desktop app for Mac and Windows that wraps the same engine in a proper interface, and that’s the version I’d point most knowledge workers to. If you’ve never touched a command line, the terminal is a hurdle you don’t need — the desktop app gives you the full thing without making you learn shell commands first.

It also runs in the web app and on your phone, with sessions syncing across all of them. I’ll start something on my laptop and check on it later from the Claude mobile app — I’ve kicked off real work from the gym between sets. Paired with a dictation app, talking to Claude Code on your phone is genuinely pleasant, as long as you’re not trying to read long outputs on a small screen.

Who should use it

Developers who want to move faster. Product managers who want to prototype without waiting on engineering. Entrepreneurs who want to build before they can afford to hire. Non-technical people who want to make small tools for their team. The bar is genuinely just this: if you can describe what you want, Claude Code can have a real go at building it. For what working with it actually feels like day to day, see my dedicated Claude Code guide.

Product 4: Claude Cowork

URL: Claude Cowork Category: AI for Productivity Full review: Claude Cowork resource page

Cowork is the newest piece, launched January 2026. Anthropic calls it “Claude Code for the rest of your work,” and that’s the cleanest way to think about it: the same agentic approach, pointed at the non-coding parts of a business.

What it does

Point Cowork at a folder on your computer, tell it what you need, and it goes off and works. It reads files, writes documents, builds spreadsheets with formulas that actually calculate, puts together presentations, and tidies up messy data — all without you walking it through every step.

The plugin ecosystem

Plugins push Cowork into specific domains: legal contract review, marketing campaigns, sales workflows, HR processes. Organizations can build their own for their own processes, and Anthropic is working toward org-wide sharing and private marketplaces.

Where it stands today

Cowork is in research preview and ships with blunt warnings about its ability to take destructive actions. It launched Max-only, but since January 16, 2026 it’s included with every paid plan — Pro at $20/month gets you in, so you no longer need the $100+ Max tier to try it. Max just buys you a bigger usage allowance (5×–20× Pro’s), not extra features. It’s powerful and not yet polished — both things are true at once. If you’re happy being an early adopter and babysitting the rough edges, it’s a real look at where knowledge work is heading.

Pricing overview

Plan	Price	What You Get
Free	$0	Claude (Sonnet), limited usage
Pro	$20/month	Claude (all models), Claude Code, Cowork, higher limits
Max	$100-200/month	Everything in Pro, substantially higher limits (5×–20× usage)
Team	$30/user/month	Pro features + team management, shared projects
Enterprise	Custom	Team features + SSO, audit logs, admin controls

My recommendation: start with Pro at $20/month. You get every model, including Opus, plus Claude Code and Cowork, at limits generous enough for most people. Max buys you more usage headroom, not more features — only move up if you keep bumping into Pro’s ceiling.

The MCP standard

Here’s a piece most guides skip. Anthropic created the Model Context Protocol (MCP), an open standard for wiring AI models up to external tools and data. It’s been picked up across the industry — even by OpenAI — and was donated to the Linux Foundation in late 2025.

Why you should care: the connectors and tools you build for Claude with MCP will most likely work with other AI platforms too. So learning this doesn’t lock you into Anthropic. You’re investing in an open standard, not a walled garden, which is exactly the kind of bet I like to make when the technology is moving this fast.

How I actually use all this

My daily stack, for what it’s worth:

• Claude (web/mobile) for conversations, brainstorming, analysis, and writing — the starting point for almost everything
• Opus specifically when I need to think hard about strategy, product planning, or a genuinely complex problem
• Claude Code for building and maintaining this website, plus other coding work
• Connectors (Google Drive, Slack, and a few others) so conversations have my real context to work with
• Perplexity plus Claude as a combo: research in Perplexity, then the deep analysis in Claude

Put together, this covers conversation, creation, coding, and autonomous work — the four things I spend most of my professional day on. If you’re serious about making AI part of how you work, this is the ecosystem I’d learn first.

Getting started

1. Sign up at claude.ai and try the free tier
2. Use it daily for a week — let it replace Google for questions, and lean on it for writing and analysis
3. Upgrade to Pro ($20/month) to unlock Opus and Claude Code
4. Connect your tools — set up Connectors for Google Drive, Slack, or whatever you live in
5. Try Claude Code — even if you don’t write code, build something small and see what happens
6. Explore Cowork — it’s included on Pro now, so hand it a messy folder and one clear instruction