beginner
Claude
Claude Pro
Claude Code
How to Use Claude: Anthropic's AI Assistant for Thoughtful Work
Master Anthropic's Claude for long-form writing, document analysis, coding, and nuanced reasoning. From your first conversation to Projects and extended thinking.
28 February 2026
claude
anthropic
platform
Create an account at claude.ai and set up your profile with role, preferences, and output style to personalise every response
Build Projects for each client or workstream with custom instructions and reference documents for persistent, contextual AI assistance
Use Artifacts to generate interactive documents, code, and visualisations that you can iterate on in a dedicated side panel
Enable extended thinking on complex tasks to get step-by-step reasoning with dramatically better results for analysis and coding
Explore Claude Code for AI-powered development directly in your terminal, with full codebase understanding
Why This Matters
Claude has established itself as the preferred AI for professionals who care about writing quality, nuanced reasoning, and safe handling of sensitive content. Built by Anthropic with a focus on being helpful, harmless, and honest, Claude consistently outperforms competitors on long-form writing, document analysis, and tasks requiring careful judgement.
What sets Claude apart isn't just capability but character. It's less likely to hallucinate confidently, more willing to say 'I don't know', and noticeably better at following complex multi-step instructions. Its 200,000 token context window means you can upload entire books, codebases, or research libraries and work with them in a single conversation.
Visit claude.ai to start using Claude today.
What sets Claude apart isn't just capability but character. It's less likely to hallucinate confidently, more willing to say 'I don't know', and noticeably better at following complex multi-step instructions. Its 200,000 token context window means you can upload entire books, codebases, or research libraries and work with them in a single conversation.
Visit claude.ai to start using Claude today.
How to Do It
1
Create your account
Go to claude.ai and sign up with your email or Google account. The free tier gives you access to Claude Sonnet with daily message limits.
2
Set up a Project
Click Projects in the sidebar and create one for your current work. Add custom instructions (your role, preferences, output style) and upload any reference files that should persist across conversations.
3
Upload your first document
Click the paperclip icon to attach files. Claude handles PDFs, code files, images, CSVs, and more. With a 200K token context window, you can upload documents that other AI tools can't process.
4
Write a specific prompt
Be direct and structured. Tell Claude its role, the task, the format, and any constraints. Example: 'You are an editor. Review this draft for clarity, conciseness, and logical flow. Be critical but constructive. Suggest specific rewrites, don't just flag issues.'
5
Enable extended thinking
For complex tasks, toggle on extended thinking (the brain icon). Claude will show its reasoning process before answering, producing significantly better results for analysis, coding, and multi-step problems.
6
Iterate within the conversation
Build on Claude's responses within the same conversation. Say 'Now rewrite section 2 to be more concise' or 'Apply that same analysis to the second document'. Context compounds across messages.
What This Actually Looks Like
The Prompt
You are a senior editor at a business publication. I'm uploading a 2,000-word draft article about AI adoption in Southeast Asian SMEs. Please: 1) Rate the article 1-10 for clarity, structure, and engagement 2) Identify the 3 weakest paragraphs and explain why 3) Rewrite the opening paragraph to hook the reader immediately 4) Suggest a better headline and 3 subheading options 5) Flag any claims that need citations Be honest and specific. I'd rather have blunt feedback than polite vagueness.
Example output — your results will vary based on your inputs
Overall rating: 6/10 - Solid research and relevant topic, but the structure buries the lead and several paragraphs repeat the same point in different words.
Three weakest paragraphs:
1. Paragraph 2 ("In today's rapidly evolving...") - Opens with a cliche and states the obvious. Your audience already knows AI is evolving. Cut entirely and move the stat from paragraph 4 here instead.
2. Paragraph 7 ("Many businesses have found...") - Vague anecdotal claims with no specifics. Replace 'many businesses' with a named example or cut it.
3. Paragraph 11 ("It is worth noting...") - 'It is worth noting' is a filler phrase. The paragraph itself makes an interesting point about regulatory gaps but takes 120 words to say what needs 40.
Rewritten opening:
A garment factory in Ho Chi Minh City cut its quality inspection time by 73% last quarter. Not by hiring more staff or buying new equipment, but by pointing a US$200 camera at its production line and connecting it to an AI model that cost less than a Netflix subscription. This isn't a Silicon Valley experiment. It's Tuesday in Southeast Asia's manufacturing sector.
Headline suggestion: "The US$200 AI Revolution Sweeping Southeast Asian Factories"
Subheadings: "Why SMEs Are Adopting Faster Than Enterprises" | "The Regulatory Gap Nobody's Talking About" | "Three Models That Actually Work at Scale"
Three weakest paragraphs:
1. Paragraph 2 ("In today's rapidly evolving...") - Opens with a cliche and states the obvious. Your audience already knows AI is evolving. Cut entirely and move the stat from paragraph 4 here instead.
2. Paragraph 7 ("Many businesses have found...") - Vague anecdotal claims with no specifics. Replace 'many businesses' with a named example or cut it.
3. Paragraph 11 ("It is worth noting...") - 'It is worth noting' is a filler phrase. The paragraph itself makes an interesting point about regulatory gaps but takes 120 words to say what needs 40.
Rewritten opening:
A garment factory in Ho Chi Minh City cut its quality inspection time by 73% last quarter. Not by hiring more staff or buying new equipment, but by pointing a US$200 camera at its production line and connecting it to an AI model that cost less than a Netflix subscription. This isn't a Silicon Valley experiment. It's Tuesday in Southeast Asia's manufacturing sector.
Headline suggestion: "The US$200 AI Revolution Sweeping Southeast Asian Factories"
Subheadings: "Why SMEs Are Adopting Faster Than Enterprises" | "The Regulatory Gap Nobody's Talking About" | "Three Models That Actually Work at Scale"
How to Edit This
Notice how the prompt asked Claude to be specific and blunt rather than generically positive. The instruction 'I'd rather have blunt feedback than polite vagueness' is key for getting honest editorial feedback. Claude naturally tends toward kindness, so being explicit about wanting directness produces much more useful output.
Prompts to Try
Deep Document Analysis
I'm uploading [DOCUMENT TYPE]. Please: 1) Provide a 3-paragraph executive summary highlighting the most important findings 2) Identify the 5 strongest arguments and 3 weakest points 3) Flag any claims that lack supporting evidence 4) Suggest 3 follow-up questions I should investigate Be direct and opinionated. Don't hedge unnecessarily.
What to expect: Claude excels at long document analysis. It will provide a structured breakdown with specific references to passages in your document. The 200K context window means it can process entire reports, contracts, or research papers in one go.
Technical Writing with Nuance
Write a [CONTENT TYPE] about [TOPIC] for [AUDIENCE]. Requirements: - Length: approximately [X] words - Tone: [professional/conversational/academic] - Include specific examples and avoid generic advice - Use British English with contractions - Structure with clear headings but keep it readable, not listy Don't start with 'In today's world' or any cliched opening. Jump straight into value.
What to expect: Claude produces noticeably more natural, less formulaic writing than other AI tools. The instruction to avoid cliches and be direct helps it lean into its strength for nuanced, human-sounding prose.
Code Review and Refactoring
Review this code and: 1) Identify bugs, edge cases, and potential security issues 2) Suggest performance improvements with explanations 3) Refactor for readability while preserving functionality 4) Add TypeScript types if missing Explain your reasoning for each change. If something is fine as-is, say so rather than suggesting changes for the sake of it. ``` [PASTE YOUR CODE] ```
What to expect: A thorough code review with specific line-by-line suggestions, not vague advice. Claude will explain trade-offs for each recommendation and won't suggest unnecessary changes just to look thorough.
Common Mistakes
Not using Projects
Starting isolated conversations instead of using Projects to maintain persistent context, instructions, and reference files across multiple chats.
Ignoring the system prompt
Not providing a system prompt or role context. Claude responds much better when you tell it who it's acting as and what standards to follow.
Wasting the context window
Uploading a 50-page PDF then asking one simple question. Use the large context window strategically for multi-document analysis, comparative work, and iterative refinement.
Being too polite in prompts
Adding excessive pleasantries and caveats instead of being direct. Claude handles direct, specific instructions better than conversational requests.
Not using extended thinking
Skipping extended thinking mode for complex reasoning tasks. Enabling it dramatically improves output for analysis, coding, and multi-step problems.
Tools That Work for This
Claude Pro
Priority access to Claude's most capable models, extended thinking, and significantly higher usage limits for US$20/month
Projects
Persistent workspaces with custom instructions, uploaded knowledge files, and conversation history that carries context across sessions
Claude API
Programmatic access for developers, with support for tool use, function calling, and batch processing
Claude for Enterprise
Team deployment with admin controls, SSO, extended context windows, and no training on your data
Getting Started: Setting Up Claude for Productive Work
Head to claude.ai and sign up with your email or Google account. The free tier gives you access to Claude Sonnet with a generous daily message allowance — more than enough to explore properly.
Claude's interface is intentionally minimal: a clean chat input, conversation history in the sidebar, and a model selector. What sets it apart immediately is its context window — Claude can process up to 200,000 tokens (roughly 150,000 words) in a single conversation, meaning you can paste entire documents, codebases, or research papers and discuss them in full.
Start by trying these:
- Paste a long document and ask Claude to summarise the key points
- Ask it to compare two approaches to a problem, with pros and cons
- Share a piece of writing and ask for detailed editorial feedback
- Upload a PDF and ask Claude to extract and organise specific information
Claude thinks before responding. Unlike tools that prioritise speed, Claude takes a moment to reason through complex questions. This means you'll often get more considered, accurate answers — especially for nuanced topics where the first-instinct answer isn't always the best one.
Claude's interface is intentionally minimal: a clean chat input, conversation history in the sidebar, and a model selector. What sets it apart immediately is its context window — Claude can process up to 200,000 tokens (roughly 150,000 words) in a single conversation, meaning you can paste entire documents, codebases, or research papers and discuss them in full.
Start by trying these:
- Paste a long document and ask Claude to summarise the key points
- Ask it to compare two approaches to a problem, with pros and cons
- Share a piece of writing and ask for detailed editorial feedback
- Upload a PDF and ask Claude to extract and organise specific information
Claude thinks before responding. Unlike tools that prioritise speed, Claude takes a moment to reason through complex questions. This means you'll often get more considered, accurate answers — especially for nuanced topics where the first-instinct answer isn't always the best one.
Projects and Artifacts: Claude's Killer Features
Beyond basic chat, Claude offers powerful features for serious work:
Projects let you create persistent workspaces with uploaded documents, custom instructions, and ongoing conversations. Perfect for research projects, client work, or any task that needs consistent context across multiple sessions.
Artifacts are interactive documents Claude creates alongside its responses — code snippets, diagrams, analyses, and documents that you can edit, iterate on, and export. When Claude writes code, it can render it as a live preview right in the conversation.
Extended Thinking (available on Claude Opus) lets Claude reason through complex problems step by step before responding. It's particularly powerful for maths, logic, coding challenges, and multi-step analysis where working through the problem methodically produces significantly better results.
Vision allows Claude to analyse images — screenshots, diagrams, handwritten notes, charts, and photos. Upload an image alongside your question for visual analysis.
Tool Use and MCP (Model Context Protocol) connects Claude to external tools and data sources, enabling it to search the web, access files, run code, and interact with other services directly within conversations.
Projects let you create persistent workspaces with uploaded documents, custom instructions, and ongoing conversations. Perfect for research projects, client work, or any task that needs consistent context across multiple sessions.
Artifacts are interactive documents Claude creates alongside its responses — code snippets, diagrams, analyses, and documents that you can edit, iterate on, and export. When Claude writes code, it can render it as a live preview right in the conversation.
Extended Thinking (available on Claude Opus) lets Claude reason through complex problems step by step before responding. It's particularly powerful for maths, logic, coding challenges, and multi-step analysis where working through the problem methodically produces significantly better results.
Vision allows Claude to analyse images — screenshots, diagrams, handwritten notes, charts, and photos. Upload an image alongside your question for visual analysis.
Tool Use and MCP (Model Context Protocol) connects Claude to external tools and data sources, enabling it to search the web, access files, run code, and interact with other services directly within conversations.
Extended Thinking, Claude Code, and Advanced Workflows
Claude responds best to clear, structured prompts with explicit context:
Give Claude a role and context. Start with: 'You are a senior data analyst reviewing quarterly sales figures for a retail chain in Southeast Asia. I'll share the data and need you to identify trends and anomalies.'
Be explicit about what you want. Claude follows instructions precisely, so specify the format, length, tone, and structure you need. If you want bullet points, say so. If you want a table, describe the columns.
Use XML tags for structure. Claude responds particularly well to structured inputs using tags like
Leverage the long context window. Don't summarise documents before sharing them with Claude — paste the full text. Claude handles long inputs remarkably well and you'll get more accurate, detailed responses from the complete source material.
Iterate with specificity. When refining Claude's output, be precise about what to change: 'In paragraph 3, replace the technical jargon with plain language suitable for a non-technical executive audience' works better than 'make it simpler'.
Give Claude a role and context. Start with: 'You are a senior data analyst reviewing quarterly sales figures for a retail chain in Southeast Asia. I'll share the data and need you to identify trends and anomalies.'
Be explicit about what you want. Claude follows instructions precisely, so specify the format, length, tone, and structure you need. If you want bullet points, say so. If you want a table, describe the columns.
Use XML tags for structure. Claude responds particularly well to structured inputs using tags like
<context>, <task>, <examples>, and <constraints>. This helps it parse complex requests accurately.Leverage the long context window. Don't summarise documents before sharing them with Claude — paste the full text. Claude handles long inputs remarkably well and you'll get more accurate, detailed responses from the complete source material.
Iterate with specificity. When refining Claude's output, be precise about what to change: 'In paragraph 3, replace the technical jargon with plain language suitable for a non-technical executive audience' works better than 'make it simpler'.
Frequently Asked Questions
Both are excellent, but they have different strengths. Claude tends to produce more nuanced, well-structured writing and excels at long-context tasks (200K tokens vs ChatGPT's 128K). ChatGPT has a larger ecosystem of plugins and GPTs, stronger image generation with DALL-E, and more third-party integrations. Many professionals use both, choosing based on the task at hand.
Claude can process up to 200,000 tokens in a single conversation - roughly equivalent to a 500-page book. This means you can upload entire codebases, lengthy reports, or multiple documents and Claude can reason across all of them simultaneously. This is particularly valuable for legal review, research synthesis, and code analysis where context across multiple files matters.
Yes. Anthropic offers Claude Team plans with centralised billing and admin controls, and Enterprise plans with SSO, audit logs, and custom data retention policies. Claude does not train on your conversations by default. Many companies across Asia-Pacific use Claude for writing, analysis, coding, and customer-facing applications via the API.
Next Steps
Set up your first Project with custom instructions for your most common type of work. Upload your style guide or brand guidelines as Project knowledge. Try extended thinking on a complex analysis task to see the quality difference. Explore Claude Code if you're a developer.