Claude Complete Guide: Shortcuts, Choosing a Model, Agents (Cowork and Claude Code) and Connectors

Published November 5, 2026 · 17 min read

Claude can do far more than answer questions in a chat window. It can work through whole tasks on your files, connect to tools like Google Drive, Slack or your own systems, and, for developers, run as a coding agent in the terminal. This tutorial shows you how to use each part: shortcuts, choosing a model, agents and connectors, first in the Claude app and then in Claude Code.

The three ways to use Claude

Part 1: Claude keyboard shortcuts

In the Claude app

• Enter sends your message and Shift + Enter starts a new line.
• Ctrl + V pastes a screenshot or image straight into the message box.
• The + button at the lower left of the message box opens files, connectors and tools.
• The desktop app has quick-entry shortcuts that open a small Claude box over any app, including one for voice. You can change or turn them off in the desktop app's Settings.

In Claude Code (the terminal agent)

Claude Code has a large set of shortcuts. These are the ones you'll use every day:

And three characters at the start of a prompt:

• / opens commands and skills, for example /model, /mcp or /agents.
• ! runs a shell command directly, for example !npm test, and Claude sees the output.
• @ mentions a file so Claude reads it, for example @src/app.js.

Tip: you can type your next message while Claude is still working. It's queued and sent when the current step finishes.

Part 2: How to set the model

The Claude models

At the time of writing, the main Claude models are:

Which ones you can pick depends on your plan.

In the Claude app

1. Look at the model name shown in the message box.
2. Click it to open the model menu and pick another model.
3. Where the model offers it, the same menu also lets you choose how much Claude thinks before answering. More thinking is slower but better for hard problems.

A good habit: start with Sonnet, and switch to Opus or Fable only when an answer isn't good enough. You'll get faster replies and use less of your plan's limits.

In Claude Code

There are several ways to set the model. The ones higher in this list win:

1. During a session: type /model to open the picker, or /model sonnet to switch straight away. In the picker, Enter switches and saves it as your default, and s switches for this session only.
2. When starting: claude --model opus
3. Environment variable: ANTHROPIC_MODEL=sonnet
4. Settings file: add "model": "opus" to your settings.json.

{
  "model": "opus"
}

Instead of full model IDs you can use short aliases:

Aliases always point to the newest version. To lock a specific version, use its full ID, such as claude-opus-5-5.

Effort: how hard Claude thinks

Effort controls how much reasoning Claude does. The levels are low, medium, high, xhigh and max. Set it with /effort (a slider), /effort high, or claude --effort high. Use lower effort for quick edits and higher effort for tricky bugs or big refactors.

Part 3: Agents in Claude

An agent is an AI that carries out a task step by step: reading files, using tools and producing a result, while you supervise. Claude has two: Cowork for everyday work, and Claude Code for software.

Cowork: give Claude a task, not a question

1. Open Claude on the web, in the desktop app or in the mobile app.
2. In the message box, select Cowork. If you don't see separate Chat and Cowork options, your account already uses the combined layout, so skip this step.
3. Describe the outcome you want, for example "Turn the invoices in my Downloads folder into a spreadsheet with date, supplier and amount."
4. Review Claude's plan, then let it run.

What Cowork can do:

• Work with your files on the desktop app, without uploading and downloading. It can only read and write in folders you've connected.
• Use a browser: open sites, read pages, click, type and fill in forms.
• Use your connectors (Part 4) such as Drive, Slack or your calendar.
• Run on a schedule: type /schedule in a task, or open Scheduled in the sidebar. Scheduled tasks run in the cloud, so your computer doesn't need to be awake.
• Produce real files: spreadsheets, presentations and formatted documents.

Choose how much Claude asks before acting, using the mode selector in the message box:

Cowork needs a paid plan (Pro, Max, Ultra, Team or Enterprise). You can set standing instructions for every session, such as your role or preferred format, in Settings > Cowork.

Claude Code subagents: your own specialist helpers

In Claude Code, subagents are specialist assistants that Claude can hand work to: a code reviewer, a test runner, a documentation writer. Each runs in its own context and reports back, which keeps your main conversation clean.

Step 1: Create the file. A subagent is a Markdown file. Save it in .claude/agents/ to share it with everyone on the project, or in ~/.claude/agents/ to use it in all your projects:

---
name: code-reviewer
description: Reviews code changes for bugs, security issues and readability. Use proactively after code is written or changed.
tools: Read, Grep, Glob
model: sonnet
---

You are a careful senior code reviewer. For each issue, explain the
problem, show the code, and suggest a fix. List the most serious issues first.

• name and description are required. Claude reads the description to decide when to use the subagent, so write it clearly. "Use proactively" encourages Claude to call it on its own.
• tools limits what it can do. This reviewer can only read, never edit.
• model sets its model: sonnet, opus, haiku, fable or inherit.
• The text below the --- lines is its instructions.

You can also simply ask Claude: "Create a code-reviewer subagent in .claude/agents that only reads files and uses Sonnet." Running /agents points you to the same folders.

Step 2: Use it. Claude delegates automatically when a task matches the description. To call it yourself:

• In plain words: "Use the code-reviewer subagent to check my last changes."
• With an @-mention, which guarantees that subagent runs: @"code-reviewer (agent)" look at the auth changes
• For a whole session: claude --agent code-reviewer

Part 4: Connectors: link Claude to your tools

Connectors let Claude read from and act in other services, such as cloud storage, calendars, project trackers and databases. They're built on the Model Context Protocol (MCP), an open standard, so any service with an MCP server can be connected.

Turn a connector on in a chat

1. Click the + button at the lower left of the message box.
2. Choose Connectors.
3. Use the switches to turn each connector on or off for this conversation.

Turning off connectors you don't need keeps Claude focused and avoids unwanted actions.

Add a custom connector (any remote MCP server)

On Pro and Max plans:

1. Go to Customize > Connectors.
2. Click + Add, then Add custom connector.
3. Enter a name and the remote MCP server URL, then click Continue.
4. Check the sign-in settings Claude detected, change them if needed, and click Continue. Add request headers such as an API key only if your server needs them.
5. Click Add, then sign in to the service if asked.

On Team and Enterprise plans, an Owner adds the connector once in Organization settings > Connectors. Members then find it under Customize > Connectors and click Connect. Free accounts can add one custom connector. To remove a connector, open Customize > Connectors and use the three-dot menu.

Connectors in Claude Code

In the terminal, add MCP servers with claude mcp add:

# A remote server (recommended)
claude mcp add --transport http notion https://mcp.notion.com/mcp

# A local server that runs on your computer
claude mcp add --env AIRTABLE_API_KEY=YOUR_KEY --transport stdio airtable -- npx -y airtable-mcp-server

# Share a server with your whole team (writes .mcp.json in the project)
claude mcp add --transport http shared-server --scope project https://example.com/mcp

Then, inside Claude Code, type /mcp to sign in to servers that need it and check their status. claude mcp list shows everything you've added and claude mcp remove notion removes one.

Connector safety checklist

• Only connect servers you trust, and read the access they ask for.
• Most connectors use a sign-in (OAuth), so Claude never sees your password. You can revoke access in Claude's settings or in the other service.
• Choose Allow always only for tools you're happy to let run without checking.
• Watch out for prompt injection: a web page or document can contain hidden instructions. Keep approvals on for anything that sends, deletes or pays.
• On team plans, an API key added to a shared connector is shared by everyone who uses it, so use a key with limited access.

Part 5: Slash commands, and how to make your own like /roadmap

In Claude Code, any message that starts with / is a command. Type / on its own to see every command available to you, or / plus a few letters to filter the list. Anything you type after the command name is passed to it, for example /plan fix the login bug.

Built-in commands worth knowing

Create your own command: /roadmap

You can turn any prompt you use often into your own slash command. Claude Code calls these skills. A skill is a folder with a SKILL.md file in it, and the folder name becomes the command.

Step 1: Create the folder. Pick where it should work:

• .claude/skills/roadmap/SKILL.md: in this project only, and shared with your team through Git.
• ~/.claude/skills/roadmap/SKILL.md: in all your projects, just for you.

Step 2: Write the file.

---
name: roadmap
description: Turn a feature idea into a phased project roadmap with milestones, tasks and risks.
argument-hint: [feature or goal]
disable-model-invocation: true
---

Create a roadmap for: $ARGUMENTS

1. Read CLAUDE.md and look at the project structure to understand the codebase.
2. Split the work into 3 to 5 phases, with the smallest useful release first.
3. For each phase, list the goal, the tasks (with the files likely to change)
   and a rough size: S, M or L.
4. End with the risks and open questions.
5. Save the roadmap to docs/ROADMAP.md and show me a short summary.

Step 3: Use it. Type the command followed by what you want planned:

/roadmap user login with Google and email

Claude replaces $ARGUMENTS with "user login with Google and email" and follows your steps. New skills work straight away, with no restart needed.

What each line means

• name: the command name. If you leave it out, the folder name is used.
• description: what the command does. Claude reads it to decide when the skill is useful.
• argument-hint: the grey hint shown while you type, here [feature or goal].
• disable-model-invocation: true: only you can run it by typing /roadmap. Leave it out if you want Claude to use the skill by itself when it fits.
• $ARGUMENTS: everything you typed after the command. For separate words, use $0, $1 and so on: /compare React Vue with "Compare $0 and $1" becomes "Compare React and Vue".

Optional extras: model: sonnet runs the command on a specific model, and allowed-tools lists tools it can use without asking each time, for example allowed-tools: Bash(git *). Older guides put commands in .claude/commands/roadmap.md. That still works, but skills are the recommended format now.

More custom command ideas

Each one is just a folder and a SKILL.md file with the steps you'd normally type out by hand.

Quick start: your first 15 minutes

1. Pick Sonnet in the model menu and ask something you'd normally search for.
2. Open + > Connectors and connect one tool you use every day.
3. Try Cowork with a small file task, in Manual mode first.
4. Developers: install Claude Code, run /model, and try Shift + Tab to plan mode before a bigger change.
5. Create one subagent, such as a code reviewer, and ask Claude to use it.
6. Turn your most-used prompt into a slash command, like /roadmap.

Using ChatGPT too? Read the matching ChatGPT guide: shortcuts, models, Work and apps.

This is an independent guide. sshot.online is not affiliated with Anthropic. Claude is a trademark of Anthropic.

Frequently asked questions

Related articles

AI Tools15 min read

ChatGPT Complete Guide: Shortcuts, Choosing a Model, Agents (Work) and Apps

Shortcuts, slash commands and your own skills like @roadmap, picking a model, delegating tasks with ChatGPT Work, and connecting apps like Google Drive.

Read More →

Screenshot Sharing8 min read

How to Share a Screenshot With ChatGPT, Claude and Other AI Tools

How to give a screenshot to ChatGPT, Claude and other AI tools: pasting and uploading, using image links with agents and APIs, prompts that get better answers, and what not to share.

Read More →

Code Sharing8 min read

How to Share Code With ChatGPT or Claude (and Get Better Answers)

Paste, upload or link code for ChatGPT and Claude the right way: keeping formatting, how much code to include, prompts that get useful answers, and what to remove before sharing.

Read More →

← All guides