Complete Claude Guide

Claude is Anthropic's AI assistant and its family of large language models.

Use it to:

• Write and rewrite content
• Research topics
• Analyze documents and data
• Summarize information
• Brainstorm ideas
• Write and debug code
• Build apps and websites
• Analyze images and PDFs
• Create documents, spreadsheets, and presentations
• Work with connected apps and external tools
• Automate multi-step workflows
• Run agentic tasks

Everyday questions

Sonnet 5

Writing and content

Sonnet 5

Fast coding

Sonnet 5

High-volume workflows

Sonnet 5

Complex reasoning

Opus 5

Advanced coding

Opus 5

Production AI agents

Opus 5

High-stakes professional work

Opus 5

Extremely difficult coding

Fable 5.1

Long-running autonomous projects

Fable 5.1

Complex, multi-stage knowledge work

Fable 5.1

Writing

Draft articles, emails, reports, scripts, posts, and documents.

Reasoning

Break hard problems into logical steps and weigh the trade-offs.

Coding

Generate, explain, debug, refactor, and review code.

Vision

Read uploaded screenshots, charts, diagrams, photos, and documents.

Data analysis

Analyze datasets, calculate metrics, write code, and build visualizations.

Document analysis

Read PDFs and uploaded files, then summarize them and pull out insights.

Web search

Find current information on the web when web access is turned on.

Research

Run multi-step research across many sources and synthesize what it finds.

Computer use

Agentic Claude products can operate computer interfaces.

Tool use

Claude models can use external tools and connected services.

Research turns Claude into a multi-step research agent that digs for you.

Projects are dedicated AI workspaces, one per area of work.

Where supported, Claude can remember useful context across conversations, such as:

• Your role
• Your projects
• People you work with
• Communication preferences
• Coding preferences
• Working style
• Ongoing project details

You repeat yourself less, because the context carries over.

Skills give Claude specialized instructions and resources for tasks you repeat.

You can build simple custom Skills from structured instructions, with no advanced programming required.

Artifacts give substantial content Claude creates its own dedicated workspace, such as:

• Documents
• Code
• Websites
• SVG graphics
• Diagrams
• Interactive components
• Apps
• Visualizations

Keep refining any Artifact through conversation.

Claude can create files such as:

• Spreadsheets
• Presentations
• Documents
• PDFs
• Charts
• Data visualizations
• Code files

Build an Excel financial model from this CSV.

From there, Claude can analyze the data, calculate metrics, build formulas, generate charts, and deliver the finished file.

Claude Code is Anthropic's agentic coding environment for developers.

Claude Code does more than generate snippets. It can work across an entire software project.

Cowork brings Claude's agentic capabilities to wider knowledge work.

With regular Claude, you collaborate through conversation. With Cowork, you delegate a job and Claude executes it.

Claude's agentic tools can act through computer interfaces. Depending on the Claude product and setup, Claude can:

• Navigate websites
• Click interface elements
• Work inside browser pages
• Open files
• Use development tools
• Run multi-step digital workflows

Claude shifts from answering questions to getting tasks done.

Claude Design is a conversational design workspace with a visual canvas.

Describe what you want, and Claude generates the design on the canvas.

Refine the result through conversation and direct edits.

Claude can connect to external tools and data.

MCP is short for Model Context Protocol.

MCP

Gives Claude access to your tools and data.

Skills

Teach Claude how to run specialized workflows.

Claude can generate visual and interactive content, such as:

• Charts
• Diagrams
• Interactive visualizations
• HTML interfaces
• SVG graphics

Claude can also analyze uploaded images. It builds visuals with code, HTML, SVG, and related technologies.

On harder tasks, Claude models can spend more computation reasoning before they answer.

Higher effort

More computation for hard problems

Lower effort

Faster, and potentially cheaper, runs

Upload your data and ask Claude to:

• Clean it
• Analyze it
• Find patterns
• Calculate statistics
• Create charts
• Write code
• Build models
• Explain findings
• Generate reports
• Create spreadsheets

Claude can help you create:

• Social posts
• Articles
• Newsletters
• Video scripts
• Hooks
• Ad copy
• Landing pages
• Emails
• Product descriptions
• Content strategies
• Research briefs

Use Claude as your personal tutor.

Teach me [TOPIC] from beginner to advanced. Use simple explanations, practical examples, exercises, and short quizzes.

Instead of: “Build an app.” Try:

Build a responsive expense-tracking web app with React and TypeScript. Write an implementation plan first. Then build the core features, test the key flows, and find and fix bugs before you show me the final result.

Instead of: “Research AI agents.” Try:

Research the current AI-agent market. Compare the main categories, companies, business models, adoption trends, and open problems. Favor primary sources and recent information. Keep facts and analysis clearly separate.

Upload any of these:

• Reports
• Research papers
• PDFs
• Contracts
• Presentations
• Spreadsheets
• Notes

Analyze this document. Return the key arguments, important numbers, assumptions, risks, contradictions, open questions, and the five most important takeaways.

Role

The role Claude should play.

Goal

Exactly what you want done.

Context

The background Claude needs.

Input

The material Claude should work from.

Constraints

The rules it must follow.

Process

How it should tackle the task.

Output

Exactly how to structure the result.

Quality check

What Claude must check before it finishes.

Act as a [ROLE]. Your goal: [GOAL]. Context: [BACKGROUND] Work from this input: [INPUT] Requirements: [REQUIREMENT 1] [REQUIREMENT 2] [REQUIREMENT 3] Constraints: [CONSTRAINTS] Process: 1. Understand the objective. 2. List the key assumptions. 3. Work through the problem step by step. 4. Verify key facts and calculations. 5. Review the result for errors. Output: [DESIRED FORMAT] Before you finish, confirm the answer is accurate, complete, logically consistent, and meets every requirement.

Learn

Teach me [TOPIC] from beginner to advanced with simple explanations and practical examples.

Research

Research [TOPIC], weigh the strongest available evidence, and give me the conclusions that matter most.

Write

Rewrite this for clarity, simplicity, and impact. Keep the original meaning.

Summarize

Boil this down to the essential ideas, key facts, decisions, and action items.

Analyze

Analyze this from several perspectives. Name the opportunities, risks, assumptions, and blind spots.

Brainstorm

Generate 30 unconventional ideas for [GOAL]. Skip the obvious ones.

Critique

Critique this constructively. Point out weak assumptions, logical gaps, and ways to improve it.

Code

Build [PROJECT]. Plan first, then implement it, test the key functionality, and fix issues before you finish.

Data

Analyze this dataset. Surface the key trends, anomalies, relationships, and actionable insights.

Decide

Compare [OPTION A] and [OPTION B] on cost, upside, downside, difficulty, time, and long-term value. Then recommend one.

Claude can still:

• Hallucinate information
• Misread ambiguous prompts
• Produce incorrect calculations
• Generate buggy code
• Misread complex documents
• Miss recent news without web access
• Make mistakes when using tools
• Fail partway through long agentic workflows

For important work, always verify the output.

Free

Best for occasional use.

Pro

Best for regular individual use.

Max

Best for heavy individual use.

Team

Best for organizations that need collaboration, admin controls, and higher usage.

Enterprise

Best for organizations that need advanced security, compliance, governance, and usage at scale.

Claude

Conversational AI assistant

Projects

Persistent workspaces

Memory

Cross-conversation context

Research

Deep multi-source investigation

Artifacts

Interactive apps and reusable deliverables

Skills

Reusable specialized workflows

Connectors + MCP

External tools and data

Claude Code

Agentic software development

Cowork

Agentic knowledge work

Claude Design

Conversational design and prototyping

API

Claude capabilities built into other products

Step 1

Pick the right Claude model.

Step 2

Give it the relevant context.

Step 3

Define the outcome you want.

Step 4

Attach files or connect data when it helps.

Step 5

Use Research for deep, current information.

Step 6

Use Projects for long-running work.

Step 7

Use Skills for workflows you repeat.

Step 8

Use Artifacts or file creation for finished deliverables.

Step 9

Use Claude Code or Cowork when Claude needs to carry out bigger tasks.

Step 10

Review and verify the end result.