A complete guide to Claude for people who run businesses.
Last verified: 13 February 2026
Claude isn't just a chatbot anymore. It's a platform — with distinct components that serve different purposes, different strengths depending on how you configure it, and a growing ecosystem of plugins and integrations. Most guides to Claude are written for developers. This one's for people who run things.
Here's everything you need to know, what's actually worth your time, and where to start.
Contents
The Platform at a Glance | Chat + Skills + Connectors | Cowork | Understanding Context | Plugins | Connectors | Claude in Chrome | Code Tab + Claude Code | Choosing a Model | Customising Claude | Where to Start | Plan Comparison | Useful Links
Claude now has five main surfaces. You don't need all of them, but understanding what each does helps you pick the right tool for the job.
Which model you get matters. Claude has three model tiers: Opus (most capable, best for complex reasoning and analysis), Sonnet (fast, good for most daily work), and Haiku (fastest, cheapest, good for simple tasks). On Pro, Claude in Chrome is limited to Haiku — you'll need Max or higher for Sonnet or Opus in the browser. In Chat and Cowork, Pro users get access to all three models.
What it is: The Claude chat window you already know — but with skills (specialist knowledge) and connectors (links to your tools) layered on top. This is where most operators spend their time, and for good reason: it's the fastest way to get work done.
Why it matters: With the right skills and connectors enabled, chat handles real work. I've been using it to build financial models in Excel, manage tasks in Asana, and draft client proposals — without leaving the conversation.
What you can actually do:
Financial modelling → Claude builds and edits .xlsx files directly, with working formulas
Task management → Claude reads from and writes to Asana, Notion, Linear, Monday.com
Document creation → Word docs, slide decks, PDFs, formatted reports
Data analysis → Query databases, build visualisations, create interactive dashboards
Email and comms → Draft and send via Gmail, Slack messages, calendar invites
The key: Enable skills (specialist knowledge) and connect your tools (via connectors). Without them, chat is a generalist. With them, it's a functional operator.
What they are: Integrations that connect Claude to your existing tools. Built on the open-source Model Context Protocol (MCP), connectors let Claude read from and write to your software — pull context from Slack, update tasks in Asana, query your database, send emails via Gmail.
The scale: The connector directory has 50+ integrations as of February 2026, spanning communication (Slack, Gmail, Microsoft 365), project management (Asana, Linear, Jira, Monday.com), content (Notion, Google Drive, WordPress), design (Canva, Figma), engineering (GitHub, Hex, Amplitude), finance (Stripe), and healthcare (Apple Health, PubMed).
Setup: Go to Settings → Connectors → browse available integrations → authenticate with your account. Most connectors use OAuth — click, approve, done.
Custom connectors: On paid plans, you can add your own MCP server URLs. If a tool has an API, you can likely connect it. This is in beta but functional.
MCP Apps: The newer addition. Some connectors now render interactive UI directly in the chat window — approve a task in Asana, send a Slack message, update a Figma component, view a Hex dashboard — without leaving the conversation. Currently live for Slack, Asana, Figma, Canva, Box, Hex, Monday.com, Clay, and Amplitude.
Availability: Directory connectors work on all plans including Free. Custom connectors require a paid plan.
What it is: Claude that acts autonomously. Cowork runs in Claude Desktop and gives Claude the ability to read your local files, browse the web, execute code, and take multi-step actions on your behalf — all in a sandboxed environment on your computer.
Why it matters: The gap between "Claude helped me think through this" and "Claude actually did this" is Cowork. It coordinates sub-agents, breaks complex work into parallel tasks, and delivers polished outputs: formatted spreadsheets, presentations, reports, processed data.
Available on: Pro, Max, Team, and Enterprise plans via Claude Desktop on macOS and Windows (x64). Windows launched 10 February 2026 with full feature parity. Not available on web or mobile.
What to know before you start:
Cowork uses significantly more of your usage allowance than standard chat. A single Cowork session can consume what several chat conversations would. Plan accordingly, especially on Pro.
Conversations are stored locally on your machine. They're not currently captured in Audit Logs, Compliance API, or Data Exports — worth knowing if you're in a regulated industry or on an Enterprise plan.
Windows arm64 (ARM-based processors) is not supported yet.
The setup that saves you hours: You can set global instructions (apply to every session) and folder-specific instructions (apply when you select that folder). Tell Claude your role, your preferences, your terminology — once. It applies everywhere. More on this in the Context section below.
This is the section most guides skip, and it's arguably the highest-leverage thing you can learn. Claude's usefulness is directly proportional to how much relevant context it has. There are several ways to give it that context — and the best ones persist between sessions.
Found in Claude Desktop settings. These apply to every Cowork session regardless of what folder you're working in.
Use them for: your role, your communication preferences, your company overview, recurring context you're tired of repeating.
Example:
I'm COO at a 15-person ecommerce company. My direct reports are Sarah (ops), Marcus (finance), and Priya (marketing). We're mid-way through migrating from Shopify to a headless setup. I prefer concise responses. Use Australian English. When creating documents, default to our brand font (Inter) and include page numbers.
Set this once. Every session starts with Claude already knowing who you are and how you work.
When you select a folder in Cowork, Claude looks for instructions specific to that folder. These add project-level context on top of your global instructions.
Use them for: project-specific terminology, file structures, approval workflows, team members involved.
Example: You have a folder for Q1 financials. The folder instructions might say: "This folder contains our Q1 FY26 close files. Key acronyms: PSR = product status report, WBR = weekly business review. The CFO wants specific numbers not ranges. Month-end close deadline is the 5th business day."
Claude can also update folder instructions on its own during a session as it learns more about the project.
This is the power move. A CLAUDE.md file is a markdown file you place in your working folder (or your home directory for global context). Claude reads it automatically at the start of every session. Think of it as the onboarding document you'd give a new chief of staff.
What to put in it:
Company overview and structure
Your role and current priorities
Team members, their roles, and how to refer to them
Key projects and their status
Acronyms and internal shorthand
File and folder structure ("reports go in /weekly-reports, client files in /clients/{name}")
Preferences and standards ("all reports start with executive summary, use metric → context → action format")
Why this matters for operators: Every Claude conversation normally starts from zero. With a well-written CLAUDE.md, Claude already knows your business, your team, and your shorthand. You can say "prep the WBR for Marcus" and Claude knows what a WBR is, who Marcus is, what format to use, and where to save it.
The Productivity plugin takes this further by creating a structured memory system: CLAUDE.md for working context, a memory/ directory for longer-term knowledge, and TASKS.md for task tracking. But you don't need the plugin to start — a single CLAUDE.md file gets you 80% of the value.
Skills are markdown files with a specific structure that give Claude specialist knowledge. They're the building blocks inside plugins, but they also work independently — you don't need a plugin to use skills.
Every skill is a SKILL.md file with two parts:
---
name: weekly-ops-report
description: Creates weekly operations reports with metrics, wins, blockers, and priorities in our standard format
---
Skills can also include scripts (Python, JavaScript, Bash) and reference files alongside the SKILL.md. They're not just passive knowledge — they can execute code.
Why this matters: Skills let you encode institutional knowledge. The stuff currently in your head, or buried in a Google Doc, or explained verbally to every new hire — that can live in a skill file that Claude follows automatically. Your report format. Your approval process. Your client onboarding checklist.
Skills are available on all plans, including Free.
What they are: Pre-built bundles that package skills, connectors, slash commands, and sub-agents together for a specific business function. Think of them as hiring a specialist who already knows the playbook for your industry.
Why they matter: Before plugins, you'd spend the first ten minutes of every session explaining your workflows. Plugins skip that. They come pre-loaded with frameworks, terminology, and standard processes.
The full stack (all eleven, open source):
| Plugin | What it does |
|---|---|
| Productivity | Tasks, persistent memory, visual dashboard. Claude remembers your team, projects, and shorthand between sessions. The /start command sets everything up. |
| Finance | Journal entries, reconciliations, variance analysis, month-end close management. Knows GAAP presentation. Can prep audit workpapers. |
| Legal | Contract review with clause-by-clause flagging, NDA triage (GREEN/YELLOW/RED risk classification), compliance checks, DPA review. |
| Sales | Prospect research, call prep, competitive intelligence, outreach drafting, pipeline review. Pulls CRM context if connected. |
| Marketing | Content drafts, campaign planning, competitive analysis, brand voice review. Handles blog posts, email sequences, social copy, SEO audits. |
| Data | SQL queries across dialects (Snowflake, BigQuery, Postgres, etc.), visualisations, interactive dashboards, statistical analysis. |
| Product Management | Feature specs/PRDs, roadmap prioritisation (RICE, MoSCoW, ICE), stakeholder comms, user research synthesis. |
| Customer Support | Ticket analysis, response drafting, escalation workflows, knowledge base management. |
| Enterprise Search | Search and retrieve information across connected enterprise tools and data sources. |
| Biology Research | Literature review, experimental design, data analysis for biological research workflows. |
| Plugin Create | The meta-plugin. Builds other plugins from your description. Also customises existing plugins for your organisation. |
How to install: Browse and install from claude.com/plugins, or in Claude Desktop go to the Cowork tab → Plugins in the left sidebar. You can also upload custom plugins directly.
All open source: github.com/anthropics/knowledge-work-plugins
What it is: A browser agent that lives in your Chrome browser. Claude can read pages, fill forms, navigate between sites, extract data, and complete multi-step browser workflows on your behalf.
Why operators should care: If you spend time on repetitive browser tasks — updating CRM records, pulling data from dashboards, filling forms, checking competitor sites — Claude in Chrome can do them for you.
Key features:
Page reading and interaction — Claude reads page content, clicks buttons, fills forms, navigates between pages
Workflow recording — Record a browser task once, and Claude learns to repeat it. Useful for any process you do the same way every time.
Scheduled tasks — Set workflows to run automatically on a schedule (daily, weekly, monthly). Check competitor pricing every Monday. Pull a report every morning.
Multi-tab management — Claude can work across multiple grouped tabs simultaneously
"Ask before acting" mode — Claude creates a plan, shows you what it intends to do, then executes only after you approve. Good for high-stakes tasks.
What to know:
Chrome only. Not supported on Edge, Brave, Arc, or any other Chromium-based browsers.
On the Pro plan, Claude in Chrome runs on Haiku 4.5 only (the smallest model). You need Max, Team, or Enterprise for Sonnet or Opus in the browser. This meaningfully limits what Pro users can do with it — Haiku handles simple tasks but struggles with complex multi-step workflows.
Install from the Chrome Web Store, then connect it to your Claude account.
What it is: Claude for building and processing. Available in two forms:
Code tab in Claude Desktop — Claude Code with a visual interface. Good for operators who want coding power without opening a terminal.
Claude Code CLI — the full command-line version.
Why operators might care: When you need to build automations, process data at scale, create custom tools, or work with files programmatically, this is where you go. The Code tab also builds plugins, so if you want to create something custom, this is often the fastest path.
Who it's for: The Code tab is for operators who want more power but prefer a UI. The CLI is for those who live in the terminal. Neither is required to get value from Claude — think of them as power tools you reach for when the standard tools aren't enough.
Claude offers three model tiers. Which one you use affects quality, speed, and how much of your usage allowance gets consumed.
Opus (most capable) — Best for complex reasoning, long documents, nuanced analysis, and tasks where accuracy matters most. Use it for: contract review, financial analysis, strategic planning, anything where you'd want a senior person's judgment. Costs the most usage.
Sonnet (balanced) — Fast and capable. Good for most daily work: drafting documents, answering questions, data analysis, content creation. The best default for operators who want quality without burning through their allowance.
Haiku (fastest) — Best for simple, high-volume tasks: quick lookups, basic formatting, simple Q&A, repetitive operations. Uses the least allowance. This is what Pro users get in Claude in Chrome.
The practical guide: Use Sonnet as your default. Switch to Opus when the task is complex or the stakes are high (a contract review, a board presentation, a financial model). Use Haiku for bulk operations or when speed matters more than depth.
Opus 4.6 — the latest — has a 200K token default context window, with up to 1M tokens available via the API for specific use cases. For most operator workflows, 200K tokens is more than enough to process substantial documents, datasets, and conversation histories.
You have a spectrum of options, from five minutes to a full afternoon:
Level 1 — Global instructions (5 minutes)
Set your role, preferences, and company context in Desktop settings. Applies to every session. Immediate payoff.
Level 2 — CLAUDE.md file (15 minutes)
Write a markdown file with your business context, team, terminology, and preferences. Place it in your working folder. Claude reads it automatically.
Level 3 — Install plugins (10 minutes)
Browse claude.com/plugins, install the ones relevant to your function. Instant specialist knowledge.
Level 4 — Customise existing plugins (30 minutes)
Use the Plugin Customiser (part of Plugin Create) to adapt Anthropic's plugins to your organisation — your terminology, your approval processes, your report formats. Describe what you need changed, Claude modifies the plugin.
Level 5 — Build your own skill or plugin (1-2 hours)
Write a SKILL.md file for a specific workflow, or ask Claude to build you a complete plugin. The Plugin Create plugin handles the structure. First attempt might need a round of feedback — customising existing plugins is often faster and gets you 80% of the way.
The honest bit: Levels 1 and 2 deliver the most value per minute invested. Most operators should start there and only move to plugins when they have a specific workflow that would benefit from structured automation.
Different Claude plans unlock different capabilities:
| Feature | Free | Pro ($20/mo) | Max ($100-200/mo) | Team ($25-30/user/mo) | Enterprise |
|---|---|---|---|---|---|
| Chat + Skills | Yes | Yes | Yes | Yes | Yes |
| Connectors (directory) | Yes | Yes | Yes | Yes | Yes |
| Custom connectors (MCP) | — | Yes | Yes | Yes | Yes |
| Cowork | — | Yes | Yes | Yes | Yes |
| Claude in Chrome | — | Haiku only | Full model access | Full model access | Full model access |
| Code Tab / Claude Code | — | Yes | Yes | Premium seats | Yes |
| Plugins | — | Yes | Yes | Yes | Yes |
| Model access | Sonnet | All three | All three | All three | All three |
| Usage | Limited | 5x Free | 20x+ Free | 1.25x Pro (Standard) / 6.25x Pro (Premium) | Custom |
Team plan note: Team plans have two seat types. Standard seats get moderate usage increases. Premium seats get significantly more usage plus Claude Code access. This matters for budgeting — not every team member needs a Premium seat.
If you're new to Claude: Start with chat. Enable a few skills relevant to your role (check Settings → Skills). Try actual work: draft a document, analyse a dataset, create a spreadsheet. Get a feel for what Claude can do in conversation before adding complexity.
Your highest-leverage next step: Write a CLAUDE.md file. Fifteen minutes documenting your role, team, and terminology will save you hours of re-explaining yourself across every future session.
When you're ready for more: Install the Productivity plugin in Cowork and run /start. It creates a structured memory system that builds on the CLAUDE.md concept with task tracking and a visual dashboard.
If you're in a specific function:
Finance → Finance plugin. Start with /reconciliation or /variance-analysis.
Legal → Legal plugin. Try /review-contract on a routine vendor agreement.
Sales → Sales plugin. Run /call-prep before your next prospect meeting.
Marketing → Marketing plugin. Good for first drafts. Always edit the output.
Data → Data plugin. Point it at a CSV and ask questions. Build a dashboard.
Product → PM plugin. Try /write-spec for your next feature.
If you want to go custom: Describe a workflow you repeat weekly to Claude. Ask it to build you a skill or plugin for it. See what happens.
Claude Desktop — Download for macOS or Windows
Plugins directory — Browse and install plugins
Connectors directory — Browse and connect your tools
Custom Skills guide — How to create your own skills
Getting started with Cowork — Official setup guide
Claude in Chrome — Setup and usage
Plugin source code — All 11 plugins, open source
MCP Protocol — Technical spec for custom connectors
Prompting guide — Anthropic's official prompting documentation
Claude for Operators is an independent publication. Not affiliated with Anthropic.
Last verified against Anthropic documentation: 13 February 2026.