A comprehensive guide to Anthropic's flagship conversational interface —
from first message to advanced power-user workflows.
claude.aiartifactsprojectsteams
Anthropic Claude • Series Deep-Dive
02
What Is Claude.ai?
Claude.ai is Anthropic's direct-to-user web application — the simplest way to interact with Claude models without writing any code.
Who It's For
Individual knowledge workers & creators
Teams needing shared AI workspaces
Enterprises requiring SSO & compliance
Anyone exploring Claude's capabilities
What It Offers
Multi-model access (Haiku, Sonnet, Opus)
Rich artifacts — code, docs, diagrams
Project-based organization & context
File uploads, images, extended thinking
Tiers
Free — limited messages | Pro ($20/mo) — 5x usage, all models | Max ($100/mo or $200/mo) — 20x usage | Team ($30/seat/mo) — shared workspaces | Enterprise — SSO, audit, custom retention
03
The Conversation Interface
Sidebar
History, search, starred chats, projects
Main Panel
Conversation thread with model selector
Artifact Panel
Live preview of generated content
04
Choosing Your Model
Use the model selector in the top-right to pick the best model for your task.
Haiku
Fastest response time
Lowest cost per message
Great for quick Q&A, brainstorms
Simple summaries & lookups
Sonnet
Best balance of speed & quality
Strong at coding & analysis
Default model for most users
Excellent for daily workflows
Opus
Most capable & nuanced
Complex reasoning & research
Long-form writing & strategy
Pro/Team/Enterprise only
05
Crafting Effective Prompts
Be Specific
Instead of "help me write code," try: "Write a Python function that takes a CSV file path and returns the top 5 rows by revenue, including error handling for missing files."
Provide Context
Share relevant background: "I'm building a Django REST API. The User model already has email and name fields. Add a profile endpoint."
Multi-Turn Refinement
Start broad, then narrow iteratively
Reference earlier messages naturally
"Make it more concise" / "Add error handling"
Claude retains full conversation context
Structured Requests
Use bullet points for multi-part asks
Specify output format (JSON, table, list)
Ask for step-by-step reasoning
Set constraints: length, tone, audience
System Prompts (Projects)
In Projects, you can set custom instructions that persist across all conversations — defining Claude's role, tone, constraints, and domain knowledge.
06
Projects
Projects let you organize conversations around a topic and give Claude persistent context.
Create
Sidebar → "New Project" → Name & describe it
Add Knowledge
Upload files or paste text that Claude references automatically
Set Instructions
Define Claude's role, constraints, and formatting preferences
07
Project Knowledge
Upload documents to give Claude persistent context across every conversation in a project.
Supported Formats
Documents: PDF, TXT, MD, DOCX, RTF
Data: CSV, JSON, XML, XLSX
Code: .py, .js, .ts, .java, .go, .rs, etc.
Images: PNG, JPG, GIF, WebP
How Context Works
Files are loaded into Claude's context window
Knowledge is available in every project conversation
Max ~200k tokens of combined knowledge
Claude can reference, quote, and reason over docs
Best Practices
Keep files focused and relevant
Use descriptive filenames
Remove boilerplate / noise
Structure text with headers
Prefer text-rich PDFs over scanned images
Text Snippets
In addition to file uploads, paste raw text as knowledge entries — useful for API docs, style guides, or reference data.
08
Artifacts
Artifacts are standalone content blocks that appear in a dedicated preview panel alongside the conversation.
Code
Python, JavaScript, HTML/CSS, SQL, and more. Syntax highlighted with copy button.
Documents
Markdown-rendered text: essays, reports, emails, plans. Full formatting support.
SVG Graphics
Vector diagrams, charts, icons, and illustrations rendered inline.
Mermaid Diagrams
Flowcharts, sequence diagrams, ER diagrams, Gantt charts rendered live.
React Components
Interactive UI components rendered in a sandboxed preview environment.
HTML Pages
Complete web pages with CSS and JavaScript, previewed in an iframe.
When Claude Creates Artifacts
Claude generates artifacts automatically when producing substantial, self-contained content. Short inline answers remain in the conversation flow.
09
Creating & Editing Artifacts
Version History
Each edit creates a new version
Navigate between versions with arrows
Compare changes across iterations
Revert to any previous version
Publishing Artifacts
Click Publish to get a shareable link
Published artifacts are hosted by Anthropic
Interactive React/HTML artifacts run live
Viewers don't need a Claude account
Editing Tips
Be specific: "Change the header color to blue" works better than "make it look nicer." Reference specific sections, variables, or line numbers for targeted edits.
10
File Uploads & Multimodal Input
Claude can analyze images, documents, data files, and code uploaded directly into conversations.
Image Analysis
Describe, interpret, and extract text from images
Analyze charts, screenshots, diagrams
Compare multiple images side-by-side
Formats: PNG, JPG, GIF, WebP
Document Processing
Summarize lengthy PDFs
Extract key data points
Answer questions about content
Cross-reference multiple documents
Data Analysis
Parse CSV/Excel files
Calculate statistics & trends
Generate charts via artifacts
Write queries against datasets
Code Review
Upload source files for review
Identify bugs and improvements
Suggest refactoring patterns
Explain unfamiliar codebases
Type
Max Size
Per Message
Notes
Images
20 MB each
Up to 5
Auto-resized for context window
PDFs
32 MB each
Up to 5
Text extraction + visual analysis
Text/Code
30 MB each
Up to 5
Most programming languages supported
CSV/Data
30 MB each
Up to 5
Parsed and analyzed structurally
11
Extended Thinking
Extended thinking gives Claude a dedicated reasoning phase before responding, visible as a collapsible "thinking" block.
When to Enable
Complex math & logic problems
Multi-step reasoning tasks
Code debugging requiring analysis
Strategic planning & decision-making
Tasks where accuracy matters most
How to Activate
Toggle "Extended thinking" in the model settings or click the lightbulb icon next to the message input.
Trade-offs
Higher accuracy on hard problems, but uses more tokens and takes longer. Keep off for simple queries.
12
Styles
Control how Claude writes using preset and custom styles applied to any conversation.
Concise
Short, direct answers. Minimal explanation. Ideal for experienced users who want quick results.
Explanatory
Detailed responses with context and reasoning. Good for learning and complex topics.
Formal
Professional tone, structured formatting. Suitable for business communications and reports.
Custom Styles
Create your own named styles
Define tone, length, formatting preferences
Paste a writing sample for Claude to match
Reuse across conversations
Share styles within Teams
Style vs. Instructions
Styles = how Claude writes (tone, length, format)
Project instructions = what Claude knows (role, context, rules)
Both can be combined for fine-tuned output
Styles persist across all conversations
Instructions are project-specific
13
Conversation Management
Organizing Conversations
Star important conversations for quick access
Rename conversations with descriptive titles
Move conversations between projects
Archive old conversations to reduce clutter
Conversations are sorted by recency by default
Search
Full-text search across all conversations
Search by title or message content
Filter by project or time period
Results show matching context snippets
Keyboard shortcut: Cmd/Ctrl + K
Continuing Conversations
Pick up any conversation where you left off. Claude retains the full context of the thread. Long conversations may hit context limits — start a new chat and summarize if needed.
Branching
Edit a previous message to branch the conversation in a new direction. The original branch is preserved — navigate between branches using the arrows above the edited message.
14
Sharing & Collaboration
Sharing Conversations
Generate a shareable link from any conversation
Recipients see the full thread (read-only)
No Claude account required to view
Revoke access at any time
Team Collaboration
Shared projects with team-wide knowledge
Team members can start new chats in shared projects
Centralized prompt libraries and templates
Activity visible to team admins
15
Keyboard Shortcuts & Power Tips
Shortcut
Action
Cmd/Ctrl + K
Search conversations
Cmd/Ctrl + Shift + N
New conversation
Cmd/Ctrl + Shift + C
Copy last response
/
Focus message input
Shift + Enter
New line (without sending)
Enter
Send message
Cmd/Ctrl + Shift + ;
Toggle sidebar
Escape
Stop generation
Power Tips
Drag & drop files anywhere in the chat
Paste images directly from clipboard
@-mention artifacts to reference them
Use Markdown in your messages for structure
Stop & retry to regenerate a response
URL Tricks
Append ?model=opus to a claude.ai URL to pre-select a model. Bookmark project URLs for quick access to specific workflows.
16
Claude.ai for Teams
$30/seat/month — Designed for collaborative workgroups who need shared context and admin controls.
Shared Workspaces
Team-wide projects with shared knowledge
Centralized prompt & style libraries
All members access the same context
Collaborative iteration on outputs
Admin Controls
Invite/remove members via email
Role management (admin, member)
Usage analytics per member
Billing centralization
Key Benefits
Higher usage limits than Pro
Early access to new features
Team data not used for training
Priority support from Anthropic
Team vs. Individual Pro
Pro gives you personal access to all models with increased limits.
Team adds shared projects, admin controls, centralized billing, and the guarantee that conversations are not used for model training.
17
Claude.ai for Enterprise
Enterprise-grade deployment with security, compliance, and governance built in.
Security & Identity
SSO/SAML — integrate with Okta, Azure AD, Google Workspace