Go back

A Guide to Claude Code for Marketers

Claude Code is Anthropic’s terminal-based AI tool that’s transforming how marketers work. Despite its name, you don’t need coding experience. You use plain English through the Terminal and get capabilities impossible with normal AI chat interfaces.

This guide covers everything marketers need to know—from installation to advanced workflows—drawing on insights from Boris Cherny (the creator of Claude Code), Andrej Karpathy, and practitioners across the industry.

Claude Artifacts: Your Day-to-Day Utility

Before diving into Claude Code, it’s worth mentioning Claude Artifacts—a feature in the standard Claude interface that’s become indispensable for daily marketing work.

Artifacts let you turn ideas into shareable apps, tools, and content by describing what you need. Claude creates standalone content in a dedicated window separate from the main conversation—making it easy to work with significant pieces of content you want to modify, build upon, or reference later.

I use Artifacts for early drafts of almost anything I publish now. They’re particularly useful for:

AI-powered Artifacts can now interact with Claude through an API—turning them into shareable apps where users can access Claude’s intelligence. When you share AI-powered artifacts, others can use them immediately with no API keys required and no costs to you.

Watch the Artifacts introduction video to see them in action.

Why Marketers Should Care

The biggest shift in 2026 is the transition from “Chat AI” (where you talk to a bot) to “Action AI” (where the bot actually does the work on your computer).

As Lenny’s Newsletter reports, the key insight is “rich context over clever prompts”—focus on providing comprehensive context about your task rather than crafting perfect prompts.

Claude Code can:

For marketers, this means analysing your entire content library, auditing hundreds of pages for SEO, generating ad variations at scale, and building presentations—all through natural language.

Getting Started

Installation

Open Terminal and run:

curl -fsSL https://claude.ai/install.sh | bash

Type claude to start. You’ll authenticate once through your browser. Requires a Claude Pro or Max subscription.

Essential Commands

Use /clear often. Every time you start something new, clear the chat—you don’t need old history consuming your tokens.

Running Shell Commands

If you need to run a command like npm run build, don’t exit Claude. Type ! followed by your command to run it directly without Claude interpreting the input.

Stopping Claude

Stopping Claude isn’t Control+C (that exits entirely). Use Escape to stop Claude mid-task.

The CLAUDE.md File

One of the most powerful features is the CLAUDE.md file—a markdown document that provides Claude with persistent context about your project, preferences, and constraints.

As Karpathy notes, this is critical for quality: “Describe code style in CLAUDE.md, describe the architecture of the relevant part, provide examples of existing similar features as templates.”

Boris Cherny’s team at Anthropic shares a single CLAUDE.md checked into git. Whenever Claude does something incorrectly, they add it to CLAUDE.md so Claude knows not to do it next time. Their current file is 2.5k tokens.

CLAUDE.md Tips from the Creator

Santiago (@svpino) recommends adding these to your CLAUDE.md:

  1. “Before writing any code, describe your approach and wait for approval. Always ask clarifying questions before writing any code if requirements are ambiguous.”

  2. “If a task requires changes to multiple files, list all files you plan to modify and get confirmation first.”

From the Claude Code team: After every correction, end with “Update your CLAUDE.md so you don’t make that mistake again.” Claude is eerily good at writing rules for itself. Ruthlessly edit your CLAUDE.md over time—keep iterating until Claude’s mistake rate measurably drops.

For marketers, your CLAUDE.md might include:

Plan Mode: The Key to Quality

One of Boris Cherny’s key techniques is starting most sessions in Plan mode.

Press Shift+Tab twice to enter Plan mode. Go back and forth with Claude until you like its plan, then switch to execution mode. Claude can often complete tasks in one shot with a good plan.

Boris estimates that 10 minutes of proper planning consistently saves an hour of debugging. Planning modes reduce AI drift and significantly improve output quality.

From the team: One person has one Claude write the plan, then spins up a second Claude to review it as a staff engineer.

Let Claude Interview You

For larger features, have Claude interview you first. As recommended in official docs, start with a minimal prompt and ask Claude to interview you using the AskUserQuestion tool:

“I want to build [brief description]. Interview me in detail using the AskUserQuestion tool. Ask about technical implementation, UI/UX, edge cases, concerns, and tradeoffs. Don’t ask obvious questions—dig into the hard parts I might not have considered. Keep interviewing until we’ve covered everything, then write a complete spec.”

Skills and Slash Commands

Skills are reusable capabilities that transform Claude from a general assistant into a specialist. They’re folders containing instructions, scripts, and resources that Claude loads when needed.

As of Claude Code 2.1.3, slash commands have merged into the skills system. A file at .claude/commands/review.md and a skill at .claude/skills/review/SKILL.md both create /review and work the same way.

From the team: “If you do something more than once a day, turn it into a skill or command.”

The Playground Skill

The playground skill generates interactive HTML explorers with controls on one side, live preview on the other, and a copy button for the generated prompt.

Templates include:

Example command: “Use the playground skill to review my CLAUDE.md and give me inline suggestions I can approve, reject, or comment on.”

Where to Find Skills

Miles Deutscher recommends bookmarking skills marketplaces—there are now 60,000+ Claude Skills ready for use.

Key sources:

Marketing-Specific Skills

Install skills that matter for marketing:

# SEO Audit
npx skills add https://github.com/coreyhaines31/marketingskills --skill seo-audit

# Copywriting
npx skills add https://github.com/coreyhaines31/marketingskills --skill copywriting

# Web Design Guidelines
npx skills add https://github.com/vercel-labs/agent-skills --skill web-design-guidelines

MCP Servers: Connecting to Your Tools

MCP (Model Context Protocol) servers give Claude direct access to external services. For marketers, this means connecting to analytics platforms, ad accounts, CMS systems, and more.

Recommended MCPs for marketers:

From the team: Enable the Slack MCP, then paste a Slack bug thread into Claude and just say “fix.” Zero context switching required.

New: You can now give Claude Code’s VS Code extension access to your browser via Claude in Chrome—enabling visual verification of your work.

Important: MCPs can consume 40%+ of context at startup. Start with 2-3 MCPs, not 10+. Use /context to see your token breakdown.

Managing Context and Costs

Context management is critical for both performance and cost. Most developers can reduce costs by 40-70% with strategic optimisation.

Key Strategies

Monitor your context: The /context command shows exactly what’s consuming your tokens. Use it regularly.

Use /compact proactively: For long sessions, use /compact to summarise the conversation. Best practice: Monitor the context meter and compact at 70% capacity.

Keep CLAUDE.md lean: Your CLAUDE.md loads at session start. Skills load on-demand only when invoked. Move specialised instructions into skills to keep your base context smaller. Aim for under ~500 lines.

Disable unused MCP servers: Each enabled MCP adds tool definitions to your system prompt. Use /mcp to disable servers not needed for your current task.

Model selection: Start with Sonnet for 80% of tasks. Only use Opus for complex analysis or architectural decisions. Not every question needs the most powerful model.

Extended thinking: Enabled by default with a 31,999 token budget. For simpler tasks, reduce costs by disabling it in /config or lowering the budget.

Cost Expectations

The average cost is $6 per developer per day, with daily costs remaining below $12 for 90% of users.

Practical Marketing Workflows

Content Library Audit

Export your content from WordPress, Webflow, or your CMS. Point Claude Code at the export:

“Analyse all articles in this folder. Create an editorial style guide identifying consistent patterns in voice, grammar, formatting, and content structure. Include examples for each pattern found.”

Claude will scan your content, identify recurring patterns, and generate a comprehensive style guide.

Transcript Analysis

As Animalz notes, Claude Code can process entire interview transcripts without compression—“you’re analysing the full content, not whatever fits the context window.”

Point Claude at a folder of transcripts:

“Analyse these interview transcripts. Extract all mentions of [topic X], identify common themes, and rank insights by relevance to our ICP.”

Growth marketers are using Claude Code to transform ad creation:

  1. Export a CSV of past ads with performance metrics
  2. Have Claude filter underperforming ads (low CTR with sufficient impressions)
  3. Generate hundreds of headline (30-char limit) and description (90-char limit) variations
  4. Export with metadata for staging

Results: Ad copy creation time went from 2 hours to 15 minutes. Creative output increased ~10x.

Key principle: “Prompting one model to manage all constraints tends to create edge-case failures.” Split tasks between focused agents—one for headlines, one for descriptions.

Interactive Presentations

Transform your analyses into self-contained HTML presentations with charts, tabs, and clickable sections—no presentation software needed to view or share.

“Create an interactive HTML presentation from this data. Include charts showing [metrics], tabs for different sections, and our brand colours.”

Writing Assistance

Claude Code works as a writing assistant: Provide context and voice instructions for drafts, then collaboratively refine line-by-line with back-and-forth feedback. Write in markdown, then paste into Notion to convert formatting for other platforms.

Advanced Techniques

Running Multiple Sessions in Parallel

Boris Cherny runs 5 Claudes in parallel in his terminal, numbered tabs 1-5, using system notifications to know when a Claude needs input. He also runs 5-10 Claudes on claude.ai in parallel.

From the team: Spin up 3-5 git worktrees at once, each running its own Claude session in parallel. It’s the single biggest productivity unlock, and the top internal tip. Some people name their worktrees and set up shell aliases (za, zb, zc) so they can hop between them in one keystroke.

You can hand off sessions between local and web using the --teleport command.

Model Selection

Boris exclusively uses Opus 4.5 with thinking for everything: “It’s the best coding model I’ve ever used, and even though it’s bigger and slower than Sonnet, since you have to steer it less and it’s better at tool use, it is almost always faster than using a smaller model in the end.”

Verification and Feedback Loops

The most important tip from the creator of Claude Code: give Claude a way to verify its work. If Claude has that feedback loop, it will 2-3x the quality of the final result.

For marketers, this might mean:

Voice Input

Talk to Claude with your voice: Use local voice transcription tools like SuperWhisper or MacWhisper to communicate faster verbally rather than typing.

Shell Aliases

Create project-specific shortcuts to accelerate your workflow. As noted in Lenny’s Newsletter, commands like cdi for loading diagrams or custom shortcuts make AI tools feel like natural extensions of your work.

Set up aliases like c for Claude, ch for Chrome integration, gb for GitHub Desktop to launch frequently-used tools quickly.

Searching Conversation History

You can ask Claude Code about your past conversations—it’ll help you find and search through them. All conversation history is stored locally in ~/.claude/, with project-specific conversations in ~/.claude/projects/.

The /insights Command

Claude Code’s /insights command (created by Alex Tamkin) analyses your past month’s history, your projects, and how you use Claude Code. It provides a project summary, usage analysis, and workflow recommendations to help you improve your productivity.

Agent Teams and Subagents

For complex tasks, Claude Code supports agent teams—multiple Claude instances working together with shared tasks, inter-agent messaging, and centralised management.

One session acts as the team lead, coordinating work, assigning tasks, and synthesising results. Teammates work independently, each in its own context window, and can communicate directly with each other.

Best use cases for agent teams:

Example prompt: “Create an agent team to review PR #142. Spawn three reviewers: one focused on security implications, one checking performance impact, one validating test coverage.”

Subagents vs Agent Teams:

SubagentsAgent Teams
CommunicationReport results back to main agent onlyTeammates message each other directly
Best forFocused tasks where only the result mattersComplex work requiring discussion and collaboration
Token costLowerHigher

Use subagents when you need quick, focused workers that report back. Use agent teams when teammates need to share findings, challenge each other, and coordinate on their own.

Agent teams are experimental—enable them by adding CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS to your settings.

Claude in PowerPoint

Claude in PowerPoint brings AI assistance directly into Microsoft PowerPoint for creating and editing presentations. It works alongside you—building slides, making pinpoint edits, and iterating on your deck in real time.

Key features for marketers:

Claude in PowerPoint is currently in beta as a research preview for Max, Team, and Enterprise plan customers.

Interactive Tools in Claude

Claude now supports interactive tools that let you open and interact with productivity apps directly within conversations—no tab switching required.

Marketing-relevant integrations:

The underlying technology is built on the Model Context Protocol (MCP). MCP Apps is a new extension that lets any MCP server deliver an interactive interface within any supporting AI product—not just Claude.

Connect to interactive apps at claude.ai/directory. Available on web and desktop for Pro, Max, Team, and Enterprise plans.

Prompting Best Practices

Research from Creator Economy shows these techniques improve Claude’s output quality:

  1. Data-first prompting — Start with your data, end with instructions. Can improve response quality by 30%.

  2. Use XML tags — Label sections clearly to prevent Claude confusing data and instructions:

    <draft>Your content here</draft>
<instructions>Your task here</instructions>

  3. Define clear roles — A clear role and task significantly improves response accuracy and writing style.

  4. Chain of thought — Ask Claude to think through complex tasks step-by-step. Proven to improve response quality by up to 39%.

File Formats as Compression

Lenny’s Newsletter highlights an underrated technique: consider which file formats best compress information for AI consumption.

“Mermaid diagrams, which are difficult for humans to parse, are perfect for machines.” Using diagram formats in markdown files can compress your application flow or content structure into a format that’s easy for AI to understand.

The Karpathy Workflow

Andrej Karpathy’s approach after extensive Claude Code use:

His 80/20 ratio flipped. He once used 80% manual work and 20% AI; now it’s 80% AI and 20% manual editing. “I really am mostly programming in English now.”

His recommended workflow: Several Claude conversation windows open on the left, working documents on the right for review and manual edits. The human handles high-quality requirements, high-level design, and final quality assurance. AI handles implementation, information retrieval, and specific tasks.

What Not to Do

Common pitfalls to avoid:

Getting Started Checklist

  1. Install Claude Code — One command, one-time browser authentication
  2. Create your CLAUDE.md — Start with brand voice and content preferences
  3. Try Plan mode — Shift+Tab twice before any significant task
  4. Install one skill — Start with SEO audit or copywriting
  5. Connect one MCP — Try GA4 or your primary analytics tool
  6. Run your first workflow — Analyse existing content or generate ad variations
  7. Run /insights — Get personalised workflow improvement suggestions

Resources

Claude Code represents a fundamental shift in how marketers will work. The terminal might feel unfamiliar at first, but the capabilities—unlimited file processing, multi-step workflows, and direct tool connections—are worth the learning curve.

Start small. Try one workflow. Build from there.

ai-and-mcp
Back to top ↑