Class Details
Claude Cowork changes how knowledge workers and teams operate. Instead of switching between tools and writing everything yourself, you configure an agent workspace that handles research, drafting, file operations, and cross-application tasks on your behalf. This class covers the full surface — from first interface orientation to running autonomous agent teams that you monitor and steer remotely.
Sixteen modules take you from the Cowork fundamentals through harness configuration, skills, connectors, MCP server design, computer use, M365 and Office add-ins, second brain knowledge architecture, multi-agent orchestration, and enterprise licensing. Each module opens with a concise theory block, then closes with a real-life business use case — so every concept lands against work you recognise on day one back at your desk.
By the end of day four you have a fully configured personal workspace, a library of custom skills, at least one working MCP connector, and a clear plan for scaling to a team or enterprise deployment. The harness patterns and skills you build during the class are production-ready and travel with you immediately.
Curriculum
Module 1 — What Cowork Is
- Chat vs. Cowork
- Folder as the agent's world
- Permissions and what Claude can access
- The sandbox and what it does not protect
- Prompt engineering for agentic tasks
- Context window & context engineering
- Claude models — when to use which
Module 2 — The Claude Desktop Interface
- Navigating the three modes: Chat, Cowork and Code
- The start screen: task input, voice input and model selector
- Starting and managing tasks: New task, Recents and Projects
- Scheduling and automation: Scheduled and Dispatch
- Observing live work: Live artifacts
- Personalising the agent: Customize
Module 3 — Claude Projects
- What a project is and how it differs from a conversation
- Project memory and how it builds over time
- Adding instructions to tailor Claude's behaviour per project
- Attaching files, PDFs and documents as project knowledge
- Connecting GitHub, Skills, Connectors and Plugins from within a project
- Research and web search inside a project
- Using style settings per project
- Launching a Cowork task directly from a project
Module 4 — Introduction to Claude Cowork
- The Cowork start screen and what it tells you
- Active tasks and returning to previous work
- Suggested tasks as a starting point
- Work in a project vs. standalone task
- Ask mode vs. Cowork mode
- Customize with plugins
Module 5 — Harness Configuration
- Structuring workspace folders
- CLAUDE.md — clear rules for Claude
- Version control your workspace and CLAUDE.md with Git
- Session directory inspection
- Debugging a failed session
- Global safety rules
- Common first-day mistakes
- Slash commands
Module 6 — Skills
- What a SKILL.md is and why it exists
- File format line by line
- Where skills live on disk and how Cowork discovers them
- Writing a skill from scratch
- Skill-creator iteration loop
- Built-in office format skills
- Cross-compatibility with Claude Code
Module 7 — Connectors & MCP
- What a connector is and how Cowork uses it
- Web connectors vs. desktop extensions
- MCP as the standard protocol
- Building a minimal MCP server
- Installing from local path vs. repository
- MCP passthrough into the Cowork session
- Testing and iterating
Module 8 — Plugins
- Plugin anatomy: skill + connector + slash command
- How plugins bundle and extend connectors
- Installing from the Plugins panel
- Building a custom plugin manifest
- Anthropic review for public plugins vs. internal plugins
- Plugin governance in an org context
Module 9 — Computer Use & Connector Hierarchy
- The fallback chain: connector → browser → screen
- Why connectors win on speed and auditability
- M365 connector as primary example
- Enabling Claude in Chrome
- When screen interaction fires
- Writing prompts that avoid unnecessary computer use
- Permission model per application
Module 10 — Office Add-ins
- Claude sidebar in Excel and PowerPoint
- Claude for Outlook
- In-file vs. cross-file decision
- Handing off from add-in to a full Cowork session
- License requirements
Module 11 — Claude Agents in M365
- Which Claude model powers each surface
- Multi-model orchestration in Copilot
- Agent 365 governance layer
- Licensing map
- Roadmap signals
Module 12 — Second Brain
- What a second brain means in an agentic context
- Structuring knowledge for Claude to find and use
- Connecting personal and organisational knowledge sources
- Notes, documents and memory as agent inputs
- Keeping the second brain current and reliable
- Deep Research
- Second brain as the foundation for automation
Module 13 — Agent Teams
- Orchestrator + sub-agents + verifier
- How Cowork decomposes a task into parallel workstreams
- Sub-agent coordination and handoff
- Observing and steering agent teams in real time
- Human-in-the-loop approval triggers
- Failure modes and recovery
- Designing tasks that benefit from parallel execution
Module 14 — Remote Claude
- Dispatch — triggering tasks without being at your desk
- Messenger integration for task control and notifications
- Scheduled tasks and recurring workflows
- Monitoring running tasks remotely
- Approving and steering tasks from mobile
- Security considerations for remote access
Module 15 — Continuous Improvements
- Treating CLAUDE.md and skills as living documents
- Reviewing session logs to identify failure patterns
- Iterating on prompts and instructions after each task
- Version controlling improvements with Git
- Building a feedback loop between users and the harness
- Measuring quality over time
Module 16 — Claude for Organisations — Team & Enterprise Plans
- Team plan vs. Enterprise plan — when each applies
- Standard seats vs. Premium seats — usage limits and pricing
- Per-member usage limits and how they reset
- SSO, Domain Capture, JIT provisioning and RBAC
- Spend controls at org and individual level
- Enterprise search across M365, Slack and custom connectors
- 150-seat ceiling and the path to Enterprise via Sales
- Admin tools, billing management and extra usage options
Class Logistics
All prices are net B2B prices and subject to 20% VAT or Reverse Charge in the EU.