Claude Code

🧠 Free Recall ❓ Quiz 🔍 Cues 🔮 Simulation 🥷 ShadowBox 📐 Encode Cards

npm install -g @anthropic-ai/claude-code

Powerful intelligence
Uses agentic search to understand your entire codebase without manual context selection
Makes coordinated changes across multiple files
Optimized specifically for code understanding and generation with Claude Opus 4

You’re in control
Never modifies your files without explicit approval
Adapts to your coding standards and patterns
Configurable; build on the SDK or run on GitHub Actions

What Claude Code does for you
Build features from descriptions: Tell Claude what you want to build in plain English. It will make a plan, write the code, and ensure it works.
Debug and fix issues: Describe a bug or paste an error message. Claude Code will analyze your codebase, identify the problem, and implement a fix.
Navigate any codebase: Ask anything about your team’s codebase, and get a thoughtful answer back. Claude Code maintains awareness of your entire project structure, can find up-to-date information from the web, and with MCP can pull from external datasources like Google Drive, Figma, and Slack.
Automate tedious tasks: Fix fiddly lint issues, resolve merge conflicts, and write release notes. Do all this in a single command from your developer machines, or automatically in CI.

Unix philosophy: Claude Code is composable and scriptable. tail -f app.log | claude -p "Slack me if you see any anomalies appear in this log stream" works. Your CI can run claude -p "If there are new text strings, translate them into French and raise a PR for @lang-fr-team to review"

Get a quick codebase overview

Start with broad questions, then narrow down to specific areas
Ask about coding conventions and patterns used in the project
Request a glossary of project-specific terms

Reference files and directories
Use @ to quickly include files or directories without waiting for Claude to read them.

Use extended thinking
Suppose you’re working on complex architectural decisions, challenging bugs, or planning multi-step implementations that require deep reasoning.

Provide context and ask Claude to think
Copy
> I need to implement a new authentication system using OAuth2 for our API. Think deeply about the best approach for implementing this in our codebase. 
Claude will gather relevant information from your codebase and use extended thinking, which will be visible in the interface.

Tips to get the most value out of extended thinking:
Extended thinking is most valuable for complex tasks such as:
Planning complex architectural changes
Debugging intricate issues
Creating implementation plans for new features
Understanding complex codebases
Evaluating tradeoffs between different approaches
The way you prompt for thinking results in varying levels of thinking depth:
“think” triggers basic extended thinking
intensifying phrases such as “think more”, “think a lot”, “think harder”, or “think longer” triggers deeper thinking
For more extended thinking prompting tips, see Extended thinking tips.

Suppose you’ve been working on a task with Claude Code and need to continue where you left off in a later session.
Claude Code provides two options for resuming previous conversations:
--continue to automatically continue the most recent conversation
--resume to display a conversation picker
Tips:
Conversation history is stored locally on your machine
Use --continue for quick access to your most recent conversation
Use --resume when you need to select a specific past conversation
When resuming, you’ll see the entire conversation history before continuing
The resumed conversation starts with the same model and configuration as the original
How it works:
Conversation Storage: All conversations are automatically saved locally with their full message history
Message Deserialization: When resuming, the entire message history is restored to maintain context
Tool State: Tool usage and results from the previous conversation are preserved
Context Restoration: The conversation resumes with all previous context intact

Run parallel Claude Code sessions with Git worktrees

Suppose you need to work on multiple tasks simultaneously with complete code isolation between Claude Code instances.

Understand Git worktrees
Git worktrees allow you to check out multiple branches from the same repository into separate directories. Each worktree has its own working directory with isolated files, while sharing the same Git history. Learn more in the official Git worktree documentation.

Tips:
Each worktree has its own independent file state, making it perfect for parallel Claude Code sessions
Changes made in one worktree won’t affect others, preventing Claude instances from interfering with each other
All worktrees share the same Git history and remote connections
For long-running tasks, you can have Claude working in one worktree while you continue development in another
Use descriptive directory names to easily identify which task each worktree is for
Remember to initialize your development environment in each new worktree according to your project’s setup. Depending on your stack, this might include:
JavaScript projects: Running dependency installation (npm install, yarn)
Python projects: Setting up virtual environments or installing with package managers
Other languages: Following your project’s standard setup process

Pipe in, pipe out

Suppose you want to pipe data into Claude, and get back data in a structured format.

Tips:
Use pipes to integrate Claude into existing shell scripts
Combine with other Unix tools for powerful workflows
Consider using —output-format for structured output

Create custom slash commands
Claude Code supports custom slash commands that you can create to quickly execute specific prompts or tasks.
For more details, see the Slash commands reference page.

Claude Code Use Cases

The Data Infrastructure team
organizes all business data for
teams across the company. They
use Claude Code for automating
routine data engineering tasks,
troubleshooting complex
infrastructure issues, and
creating documented workflows
for technical and non-technical
team members to access and
manipulate data independently.

The team showed finance team members how to write plain text files describing their data workflows, then load them into Claude Code to get fully automated execution. Employees with no coding experience could describe steps like “query this dashboard, get information, run these queries, produce Excel output,” and Claude Code would execute the entire workflow, including asking for required inputs like dates.

The team asks Claude Code to summarize completed work sessions and suggest improvements at the end of each task. This creates a continuous improvement loop where Claude Code helps refine the Claude.md documentation and workflow instructions based on actual usage, making subsequent iterations more effective.

When working on long-running data tasks, they open multiple instances of Claude Code in different repositories for different projects. Each instance maintains full context, so when they switch back after hours or days, Claude Code remembers exactly what they were doing and where they left off, enabling true parallel workflow management without context loss.

Tips from Claude Team
The better you document your workflows, tools, and expectations in Claude.md files, the better Claude Code performs. This made Claude Code excel at routine tasks like setting up new data pipelines when you have existing patterns.

They recommend using MCP servers rather than the BigQuery CLI to maintain better security control over what Claude Code can access, especially for handling sensitive data that requires logging or has potential privacy concerns.

The team held sessions where members demonstrated their Claude Code workflows to each other. This helped spread best practices and showed different ways to use the tool they might not have discovered on their own.

Set up Claude to verify its own work by running builds, tests, and lints automatically. This allows Claude to work longer autonomously and catch its own mistakes, especially effective when you ask Claude to generate tests before writing code.

When components have similar names or functions, be extremely specific in your requests. The better and more detailed your prompt, the more you can trust Claude to work independently without unexpected changes to the wrong parts of the codebase

Security engineering uses 50% of all custom slash command implementations in the entire monorepo. These custom commands streamline specific workflows and speed up repeated tasks.

Instead of asking targeted questions for code snippets, they now tell Claude Code to “commit your work as you go” and let it work autonomously with periodic check-ins, resulting in more comprehensive solutions.

Beyond coding, Claude Code excels at synthesizing documentation and creating structured outputs. They provide writing samples and formatting preferences to get documents they can immediately use in Slack, Google Docs, and other tools to avoid interface switching fatigue.

After writing core functionality, they ask Claude to write comprehensive unit tests. Claude automatically includes missed edge cases, completing what would normally take significant mental energy in minutes, acting like a coding assistant they can review

Try asking various questions to see if Claude can answer faster than Google search. If it’s faster and more accurate, it’s a valuable time-saving tool for your workflow.

Give Claude specific instructions and ask it to write logic, then verify correctness. This helps build trust in the tool’s capabilities before using it for more complex tasks

Having Claude write unit tests relieves significant pressure from daily development work. Leverage this feature to maintain code quality without spending time thinking through all test cases manually.

Save your state before letting Claude work, let it run for 30 minutes, then either accept the result or start fresh rather than trying to wrestle with corrections. Starting over often has a higher success rate than trying to fix Claude’s mistakes

While supervising, don't hesitate to stop Claude and ask "why are you doing this? Try something simpler." The model tends toward more complex solutions by default but responds well to requests for simpler approaches.

Begin with just the bare minimum of what you need and let Claude guide you through the process, rather than front-loading extensive explanations

Look for workflows involving repetitive actions with tools that have APIs (like ad platforms, design tools, analytics platforms). These are prime candidates for automation and where Claude Code provides the most value.

Instead of trying to handle everything in one prompt or workflow, create separate agents for specific tasks (like their headline agent vs. description agent). This makes debugging easier and improves output quality when dealing with complex requirements.

Spend significant time upfront using Claude.ai to think through your entire workflow, then have Claude.ai create a comprehensive prompt and code structure for Claude Code to reference. Also, work step-by-step rather than asking for one-shot solutions to avoid Claude getting overwhelmed by complex tasks.

How Marketing Uses Claude Code

The team built an agentic workflow that processes CSV files containing hundreds of existing ads with performance metrics, identifies underperforming ads for iteration, and generates new variations that meet strict character limits (30 characters for headlines, 90 for descriptions). Using two specialized sub-agents (one for headlines, one for descriptions), the system can generate hundreds of new ads in minutes instead of requiring manual creation across multiple campaigns. This has enabled them to test and iterate at scale, something that would have taken a significant amount of time to achieve previously.

Instead of manually duplicating and editing static images for paid social ads, they developed a Figma plugin that identifies frames and programmatically generates up to 100 ad variations by swapping headlines and descriptions, reducing what would take hours of copypasting to half a second per batch. This enables l0x creative output, allowing the team to test vastly more creative variations across key social channels.

They created an MCP server integrated with Meta Ads API to query campaign performance, spending data, and ad effectiveness directly within the Claude Desktop app, eliminating the need to switch between platforms for performance analysis, saving critical time where every efficiency gain translates to better ROI.

They implemented a rudimentary memory system that logs hypotheses and experiments across ad iterations, allowing the system to pull previous test results into context when generating new variations, creating a selfimproving testing framework. This enables systematic experimentation that would be impossible to track manually.

Use Command+V to paste screenshots directly into Claude Code - it excels at reading designs and generating functional code, making it invaluable for turning static mockups into interactive prototypes that engineers can immediately understand and build upon.

Add instructions to your Claude.md file to prevent Claude from making repeated tool-calling mistakes, such as telling it to “run pytest not run and don’t cd unnecessarily - just use the right path.” This significantly improved consistency.

Regularly commit your work as Claude makes changes so you can easily roll back when experiments don't work out. This enables a more experimental approach to development without risk.

Give Claude a quick prompt and let it attempt the full implementation first. Ifit works (about one-third of the time), you've saved significant time. If not, then switch to a more collaborative, guided approach.

MCPs

claude mcp add --transport http context7 https://mcp.context7.com/mcp

What is an MCP Server?
MCP (Model Context Protocol) server is a standardized way to extend AI models with additional capabilities and context. Think of it as a plugin or extension system for AI assistants.
An MCP server can provide:
Tools - Functions the AI can call (like file operations, API calls, calculations)
Resources - Access to data sources (databases, files, web content)
Context - Additional information to help the AI understand your specific domain/project
For example, an MCP server might:
Give Claude access to your company's internal documentation
Provide tools to interact with your specific APIs
Add domain-specific knowledge (legal, medical, etc.)
Connect to external services (databases, cloud platforms)
So when you run that claude mcp add command, you're essentially installing a "plugin" that gives Claude additional superpowers specific to whatever Context7 provides.

MCP Client (like Claude Code)
Role: Consumes capabilities from MCP servers Infrastructure:
Lightweight application that makes requests
Maintains connections to one or more MCP servers
Handles user interactions and translates them into MCP protocol calls
Manages authentication and connection lifecycle
Examples: Claude Code CLI, other AI tools that support MCP
What it runs:
Simple executable/application
Minimal resource requirements
Usually runs locally on developer machine
MCP Server
Role: Provides capabilities, tools, and resources Infrastructure varies significantly based on what it does:
Simple MCP Server
Just a small application listening on a port
Might be a Python/Node.js script with a few endpoints
Could run on a single container or small VM
Minimal database (maybe just local files)
Complex MCP Server (like Context7)
Web servers handling HTTP requests
Databases storing context, embeddings, or user data
Background processing for indexing, ML inference
External integrations to other APIs/services
Load balancers for scaling
Authentication systems
Caching layers (Redis, etc.)
Message queues for async processing

Unstarred

Dunk (exclude from study)

Claude Code

Review Questions

Todos

Sensemaker