Overview In this two-day course, students will be introduced to using Claude Cowork, which is Anthropic's autonomous AI agent language and execution system used to manage complex, multi-step knowledge work directly within the Claude Desktop (https://claude.ai) environment.
Description
Overview
In this two-day course, students will be introduced to using Claude Cowork, which is Anthropic's autonomous AI agent language and execution system used to manage complex, multi-step knowledge work directly within the Claude Desktop (https://claude.ai) environment. The course covers the structural requirements, permission frameworks, and interface dynamics needed to safely run computer use capabilities, but the emphasis is on leveraging Claude Cowork to handle long-running, autonomous business workflows. Students will learn how to configure persistent workspace context files, integrate native connectors and third-party Model Context Protocol (MCP) servers, build reusable custom AI skills, and schedule fully automated daily or weekly tasks. This instructor-led course consists of a combination of lecture, demonstrations, and hands-on labs.
Software Used During the Class
The standard lab setup for this class consists of a Windows or macOS PC with the Claude Desktop App (with Cowork mode enabled) installed, a modern text or Markdown editor (such as VS Code or Obsidian), a web browser with the Claude extension, and an environment preconfigured with target cloud tools (such as Google Workspace).
Audience
This course is intended for business users, operations managers, and professional knowledge workers who have been using standard conversational interfaces like Claude Chat and are ready to step into the power of autonomous AI agents. Delegating multi-step, background-running workflows is a highly effective next step for experienced Claude users looking to scale their productivity.
Prerequisites
Although deep technical coding experience is not required for this class, students should have prior experience interacting with Claude Chat or similar large language models. Basic familiarity with navigating local file structures and utilizing standard markdown formatting will help students get the most out of the hands-on lab work.
Course Outline
Day 1: Foundations, Workspace Architecture, and Browser Automation
Introduction to the Claude Agentic Ecosystem
Understanding the AI spectrum: Claude Chat vs. Claude Code vs. Claude Cowork
Subscription requirements, enterprise licensing tiers, and usage limits
Core architecture of autonomous agents: Planning, tool execution, and self-correction
Exploring the parallel background execution engine and persistent artifacts
Environment Setup, Security, and Desktop Controls
Installing and enabling Cowork mode within the Claude Desktop interface
Digital hygiene: Granting safe folder permissions and local file system access
Setting up sandboxed boundaries to prevent unintended system actions
Executing your first basic multi-step file manipulation task
Workspace Management, Context Control, and "Business Brain" Configuration
The structural importance of persistent workspace memory
Creating and optimizing the CLAUDE.md and MEMORY.md directory architecture
Implementing the "300-Line Rule" to prevent context degradation and memory drift
Establishing global instruction files and a master "Business Brain" root profile
Session pruning, archiving legacy conversations, and data cleanup strategies
Migrating a standard conversational Claude Project into a standalone Cowork workstation
Browser Use, Web Research, and Background Workflows
How Claude navigates the live web using background Chromium automation
Initiating deep market research, competitive analysis, and due diligence sessions
Leveraging the Claude in Chrome browser extension for real-time manual data gathering
Overcoming HTML scraping hurdles to extract clean, structured data from complex sites
Monitoring parallel background research tracks while working on primary tasks
Executing an autonomous, end-to-end SaaS market analysis and compiling a structured findings report
Day 2: MCP Integrations, Custom Skills, and Scheduled Automations
Expanding Agent Capabilities with Connectors and MCP
Overview of native cloud Connectors (Gmail, Google Calendar, Google Drive, Notion)
Authenticating, authorizing, and securing external application data channels
Introduction to the open-source Model Context Protocol (MCP) architecture
Connecting advanced third-party developer tools (e.g., YouTube transcript extraction via Composio MCP)
Integrating productivity webhooks and software dashboards (e.g., Fireflies meeting notes, platform metrics)
Chaining cross-application tool workflows to eliminate manual copy-pasting
Building Custom Reusable AI Skills
Differentiating between a general Cowork Workstation and a specialized, reusable Skill
Writing explicit, high-signal instruction templates and prompt definitions for custom skills
Building an automated Document-to-Presentation Skill
Designing an Infographic and Visual Asset Generation Skill
Content repurposing frameworks: Transforming raw long-form audio/video into structured social media
Financial and administrative automation: Expense report sorting, receipt parsing, and data visualization
Testing, versioning, and refining custom skills with evaluation examples
Reviewing and utilizing persistent artifact outputs effectively
Scheduled Tasks, Dispatch, and Production Workflows
Introduction to Scheduled Tasks and time-driven autonomous automations
Configuring autonomous daily triggers (e.g., building a "Morning Brief" pipeline)
Utilizing the "Dispatch" feature to chain multiple background actions
Managing token usage, monitoring tracking data, and optimizing execution costs
Troubleshooting runtime errors and gracefully managing permission popup interruptions
Sharing skills, workspaces, and automations across teams
Developing an end-to-end autonomous business workflow from the ground up
Course capstone: Designing, deploying, and validating a production-ready autonomous workflow
