Claude.ai — the hosted chat product, plainly
What Claude.ai is, the surfaces it ships on (web · desktop · iOS), what the tiers actually unlock, and how it relates to Claude Code, the API, and the Anthropic Console. Plus the privacy posture in plain terms.
The thirty-second version#
Claude.ai is Anthropic’s hosted chat product — the place you go in a browser (or desktop app, or on your phone) to talk to Claude. Same model family as the API. Different interface, different feature set, different pricing.
- Web app — the canonical surface, at claude.ai
- Desktop apps — macOS (Intel + Apple Silicon) and Windows (x64 + ARM64), same UI as the web app plus integrations
- iOS app — mobile chat, voice mode, image input
- Android app — mobile chat for the Android side of the fence
All four sync your conversation history, your Projects, and your account settings. Pick whichever fits the device.
What you can do that’s not in the API#
The API is “send a message, get a response.” Claude.ai is the API plus a bunch of product features:
| Feature | What it adds |
|---|---|
| Projects | Long-lived context — drop files, set instructions, keep coming back to the same conversation surface |
| Skills | Markdown instruction folders Claude auto-loads on relevant tasks — paid tiers only |
| Artifacts | Auto-generated side panels for code, docs, diagrams · live-edits as you iterate |
| File uploads | Drop PDFs, spreadsheets, images directly · Claude reads them inline |
| Computer Use (beta) | The agent controls a virtual computer · still beta, narrow use cases |
| Voice mode | Talk to Claude verbally · transcript shows in chat |
| Web search | When relevant · grounds Claude’s answers in current pages |
See §WEB.2 Projects and §WEB.3 Skills for the two features worth a deep dive.
The tier matrix#
Pricing as of 14 May 2026 — always verify against anthropic.com/pricing before committing. Prices in USD.
| Tier | Cost | Best for | What changes |
|---|---|---|---|
| Free | $0 | Casual use | Limited messages/day · access to a smaller model set · no Projects · no Skills |
| Pro | $17/mo billed annually ($200 up front) · $20/mo billed monthly | Daily individual use | Higher message budget · Sonnet + Opus · Projects · Skills · longer context · includes Claude Code |
| Max | $100/mo (5× Pro) or $200/mo (20× Pro) | Heavy individual use | 5× or 20× the Pro usage · priority access during peak times · more Computer Use |
| Team | $20/seat/mo annual · $25/seat/mo monthly · 5–150 seats | Small to mid teams | Pro features per user · shared billing · admin controls · workspace Projects + Skills · audit logs |
| Enterprise | Custom | Larger orgs | SSO/SAML · SCIM · custom data retention · dedicated support · contracts |
The non-obvious thing: paid tiers don’t change the model; they change the budget. Pro isn’t “smarter Claude” than Free — it’s more Claude per day.
Free vs Pro — the practical difference#
The Free tier resets quickly enough for occasional questions. Where Pro pays off:
- You use it as a daily-driver assistant
- You upload files often (Free has tighter file size limits)
- You need Projects (Free can’t create them)
- You need Skills (paid only)
- You want the option to switch to Opus on hard problems
- You also want Claude Code without a separate subscription
Don’t pay for Pro to “try Claude.” Pay for it once Free is consistently limiting you.
Where it sits vs the rest of Anthropic’s surfaces#
| If you want to… | Use |
|---|---|
| Chat through a polished UI | Claude.ai |
| Code in your terminal with agentic loops | Claude Code (auth with your Claude.ai subscription) |
| Build your own app on top of Claude | Claude API |
| Make Claude controllable by other AI tools | MCP server (talks to Claude.ai + Claude Code + others) |
Important: Pro / Max / Team / Enterprise subscriptions include Claude Code access. You log into the CLI with the same account; usage draws from your subscription’s budget rather than billing your card separately. Free does NOT include Claude Code.
Privacy in plain terms#
The privacy story is the question people ask most often. Reading privacy.claude.com carefully:
- By default, Anthropic does NOT train on your conversations for Free, Pro, Max, Team, or Enterprise — unless you opt in (verify current policy on the privacy site, this is subject to change).
- Retention windows differ by surface:
- Claude.ai (Free / Pro / Max): conversations persist until you delete them. Deletion is at your control.
- API / commercial use: inputs and outputs are deleted within 30 days of receipt.
- Flagged content (any tier): if automated safety detection flags a conversation, inputs/outputs are retained up to 2 years, plus trust-and-safety scores for up to 7 years, regardless of tier.
- Trust-and-safety review is automated first, with humans reviewing flagged content if needed. This is policy across all tiers.
- You control deletion from your own history any time. What’s retained server-side after that follows the rules above.
⚠️ What Anthropic publishes vs what they could change. This is current policy, not a technical guarantee. If your data sensitivity makes that uncomfortable, route Claude through AWS Bedrock or Google Vertex (the data stays in your cloud account under that cloud’s privacy regime). Self-hosted Claude is not generally available — it’s contract-specific Enterprise territory at best; don’t assume it exists for your use case without confirming with Anthropic Sales.
Honest takes#
What it’s genuinely good at:
- Long, file-heavy projects via Projects (worth comparing against ChatGPT Projects on your own workflow)
- Multi-step reasoning when you give it room to think
- Code review and analysis when you can drop the codebase as files
- Artifacts — the side-panel rendering for docs/code/diagrams is consistently useful
What’s frustrating:
- File upload limits feel tight at the Pro tier when working with many docs
- Voice mode is improving but the gap vs ChatGPT’s voice mode is still real
- Computer Use is genuinely beta — most tasks fail more than they succeed
What you’d switch to if not Claude.ai:
- ChatGPT (OpenAI) — similar shape, different model strengths, broader feature surface today
- Gemini chat (Google) — best at very-long-document handling + the audio overview workflow via NotebookLM
- Self-hosted via API — when privacy/data needs override polish
What to do next#
- §WEB.2 Projects feature — the high-value paid-tier feature
- §WEB.3 Skills feature — the custom-tool extension surface
- §CC.1 Claude Code overview — the same engine for terminal-shaped work
- §API.1 Claude API overview — the same engine for build-your-own apps