The Claude
Platform Playbook

Four Interfaces. One Operating System.

Claude is not a chatbot. Across four distinct interfaces — Chat, Power Features, Claude Code, and Cowork — it functions as a complete work operating system for the teams that know how to use it. Most organizations use roughly 20% of what is available to them. This playbook closes that gap.

Each section covers the core capabilities of one interface, the highest-leverage workflows for business operations, prompt patterns that produce consistent results, and the failure modes to avoid.

Claude Chat is the primary conversational interface available at claude.ai. It is session-based: each conversation has its own context window. Unless Projects or Memory is enabled, Claude does not carry knowledge between sessions. The context window — up to 1M tokens on Sonnet and Opus — handles an entire project brief, multiple documents, and extensive supporting materials in a single session.

What Chat Can Do

Three Patterns That Produce Consistent Results

These are reusable structures. Treat them as starting frames, not rigid scripts.

High-Value Workflows in Chat

What Goes Wrong in Chat

Power features transform Claude from a per-session chatbot into a persistent work environment. They are available on Pro and Max plans. The six features below represent the highest-leverage capabilities for business operations teams.

Projects Persistent workspaces with document memory

Projects are persistent workspaces. Files, instructions, and conversation history persist across sessions. Each Project has its own memory scope, preventing context from one engagement leaking into another.

1. Create a Project for each client engagement, service line, or internal workstream (e.g., "Client: Acme Corp", "Proposal Templates", "Operations").
2. Upload foundational documents: discovery notes, signed SOWs, intake forms, existing proposals, process documentation.
3. Set Project Instructions to define context, tone, and constraints that apply to all conversations in that workspace.
4. Use all subsequent conversations within that Project. Claude references uploaded files and maintains your instruction context automatically.

Memory Cross-session context retention

Memory allows Claude to retain information derived from past conversations. It builds a profile across sessions covering your role, ongoing projects, communication preferences, and recurring context. This eliminates re-establishing context at the start of every chat.

• Claude generates memory summaries from chat history and stores them by domain: Role and Work, Current Projects, Preferences and Constraints.
• Memory is separate from Projects. Project memory is document-based. Conversation memory is AI-derived and synthesized from past interactions.
• View, edit, and delete individual memories from Settings. Use this to correct any inaccuracies Claude has inferred.
• Memory is disabled in Incognito conversations. Use Incognito when working on sensitive information you do not want retained.
• Available on Pro, Max, Team, and Enterprise plans.

Deep Research Multi-step, multi-source synthesis

Deep Research is a multi-step research mode where Claude coordinates multiple search queries, reads full web pages, synthesizes across sources, and produces a structured report with citations. It is significantly more thorough than a standard web search response.

When to use Deep Research:

• Competitive landscape analysis before entering a new vertical or responding to an RFP
• Market sizing for a prospect's industry before a discovery call
• Regulatory and compliance context for industries you are targeting
• Technology comparison across multiple platforms for a specific use case
• Grant or funding program eligibility research on behalf of clients

Extended Thinking Deeper reasoning on complex problems

Extended Thinking allocates additional reasoning budget to Claude before it produces a response. Instead of answering immediately, Claude works through the problem internally before delivering a final output. The result is materially better on complex, multi-variable decisions. Now available on all three current models: Opus 4.6, Sonnet 4.6, and Haiku 4.5.

When to activate Extended Thinking:

• Evaluating whether a business process is a strong candidate for AI automation
• Sequencing multiple AI use cases across an implementation roadmap
• Pricing complex custom engagements where scope is ambiguous
• Diagnosing why an automated workflow is producing unexpected results
• Reviewing a contract or SOW for gaps and risk exposure

MCP Integrations Live connections to your tools

MCP (Model Context Protocol) allows Claude to connect directly to external tools and services. Instead of copy-pasting data between platforms, Claude reads from and writes to them live. The integration ecosystem continues to expand — below are key integrations available on claude.ai and across Claude interfaces.

Artifacts Live rendered outputs in a side panel

Artifacts open a live side-panel where Claude renders interactive content including code, HTML pages, React components, SVG diagrams, and markdown documents. You can edit and iterate in real time without switching tools.

High-value Artifact types for business teams:

• HTML pages — Client-facing proposal summaries, service one-pagers, event landing pages.
• React components — Interactive ROI calculators, assessment tools, implementation progress trackers.
• Mermaid diagrams — Workflow maps, CRM pipeline visualizations, process documentation for client handoffs.
• Markdown documents — Proposal drafts, training curriculum outlines, SOPs that export cleanly to other formats.

Claude Code is Anthropic's agentic coding tool. It operates in your terminal, IDE, desktop app, or browser — understanding your codebase and executing tasks from natural language instructions. It reads and writes files, runs bash commands, manages git workflows, connects to external services via MCP, and supports scheduled tasks, plugins, and remote control from any device.

It is the right tool for building AI agents, automating repetitive file and data operations, generating deliverables programmatically, and managing the technical infrastructure behind your AI workflows.

Requirements and Setup

• A Claude Pro, Max, Team, or Enterprise subscription, or an API key from the Anthropic Console
• macOS, Linux, or Windows (Git for Windows required on Windows)
• Multiple installation methods: native installer (auto-updates), Homebrew, or WinGet

Available across five surfaces — all connected to the same underlying engine. Your CLAUDE.md files, settings, and MCP servers work across all of them:

Core Slash Commands and Flags

CLAUDE.md and Auto Memory — Persistent Project Intelligence

Claude Code has two complementary memory systems. CLAUDE.md files are instructions you write — loaded at the start of every session. Auto memory is notes Claude writes itself based on your corrections and preferences, stored in ~/.claude/projects/. Together, they eliminate re-establishing context between sessions.

CLAUDE.md files can live at multiple levels — project root, user home (~/.claude/CLAUDE.md), or organization-managed. More specific locations take precedence. Use @path/to/file syntax to import additional files into your CLAUDE.md.

Auto Memory

Auto memory lets Claude accumulate knowledge across sessions without manual effort. Claude saves build commands, debugging insights, architecture patterns, and your preferences automatically. Run /memory to browse, edit, or delete what Claude has stored. The first 200 lines of the memory index are loaded at the start of every session.

.claude/rules/ — Path-Scoped Instructions

For larger projects, organize instructions into .claude/rules/ as separate markdown files. Rules can be scoped to specific file paths using YAML frontmatter, so they only load when Claude works with matching files. This keeps context lean and relevant.

Agent Teams and Sub-agents

Spawn multiple Claude Code agents that work on different parts of a task simultaneously. A lead agent coordinates the work, assigns subtasks, and merges results. Agent teams can run in parallel using git worktrees for isolation. Define custom agents via CLAUDE.md or the --agents flag.

Skills and Custom Commands

Create reusable workflows your team can share. Package repeatable tasks as slash commands like /review-pr or /deploy-staging. Skills load on demand — only when invoked — keeping context clean. Place skill files in .claude/commands/ for project-wide commands or ~/.claude/commands/ for personal commands.

Hooks

Run shell commands automatically before or after Claude Code actions. Auto-format files after every edit, run lint before commits, or trigger custom scripts tied to Claude's workflow. Hooks give you programmatic control over Claude's behavior at defined lifecycle points.

Scheduled Tasks

Run Claude on a recurring schedule: morning PR reviews, overnight CI failure analysis, weekly dependency audits, or syncing docs after PRs merge. Cloud scheduled tasks run on Anthropic infrastructure and keep running when your computer is off. Desktop scheduled tasks run on your machine with direct local file access. Create them from the web, the Desktop app, or by running /schedule in the CLI.

Remote Control and Cross-Device

Start a session in your terminal and continue from your phone or browser via Remote Control. Hand off terminal sessions to the Desktop app with /desktop for visual diff review. Push events from Telegram, Discord, or webhooks into active sessions via Channels. Mention @Claude in Slack with a bug report and get a pull request back.

Plugins

Install pre-built plugin bundles that add MCP servers, skills, and tools. Manage plugins with claude plugin install and browse available options from plugin marketplaces. Plugins extend Claude Code's capabilities without manual MCP server configuration.

Checkpoints and Rewind

Claude Code automatically saves checkpoints before each code change. If a batch of edits goes in an unwanted direction, press Escape twice or type /rewind to roll back to a prior state. You can restore the code, the conversation, or both independently.

High-Value Workflows in Claude Code

What Goes Wrong in Claude Code

Cowork is a research preview feature in the Claude Desktop app that brings agentic capabilities to knowledge work. Unlike Chat, which responds to individual prompts, Cowork handles complex multi-step tasks autonomously — reading and writing local files, coordinating sub-agents in parallel, and producing professional documents directly to your file system.

Available on Pro, Max, Team, and Enterprise plans. Requires the Claude Desktop app on macOS or Windows with an active internet connection. The desktop app must remain open during task execution.

What Cowork Can Do

Cowork Execution Flow

1. Describe your task in natural language. Provide any relevant files by selecting a workspace folder or uploading documents.
2. Claude analyzes the request, creates a plan, and breaks the work into subtasks. You maintain visibility throughout and can provide mid-task direction.
3. Claude executes code in an isolated virtual machine, coordinates parallel workstreams, and delivers outputs directly to your file system.
4. Review the completed deliverables in your workspace folder. Iterate with follow-up instructions if needed.

High-Value Workflows in Cowork

What Goes Wrong in Cowork

Which Model to Use

Claude Code and the API support three current models. Choosing the right one affects output quality, speed, and cost. Sonnet is the right default for most work.