Going further

Further reading and glossary

You do not need any of this to start. Bookmark it for the second week. The reading list below is the short list — the writers and posts that materially shaped this guide. The glossary disambiguates the words around Cowork that get mixed up in conversation. And the feedback channel is where you tell us what is missing.

§1 — Reading list

Recommended reading

Most Cowork material on the open web is shallow — “ten things to try” lists and screenshot tours. These six are the ones worth your attention.

§2 — Glossary

Glossary

When pilot participants disagree about something, it is usually because two of the words below are being used for the same thing, or one of them is being used for two different things. Pin these down once.

Cowork
The agentic mode inside Claude Desktop. Reads and writes a folder you connect; works on outcomes you describe rather than step-by-step instructions. The subject of this guide. Went generally available 9 April 2026.
Cowork Project
A single folder connected to Cowork, with its own Instructions field, its own memory, and its own scheduled tasks. You can have many. They do not share memory.
Instructions field
The text box in Cowork where you paste your INSTRUCTIONS.md contents. Claude reads it at the start of every session in that project. The full template lives on the Folder page.
MEMORY.md
The running whiteboard. A file in the root of your Cowork folder where Claude appends short, dated lessons it has learned. Capped at 150 lines; overflows to ARCHIVE.md.
ARCHIVE.md
The overflow file for MEMORY.md. Claude does not read it unless you ask. Lets you keep the running whiteboard small without losing the history.
Context folder
The context/ subfolder inside your Cowork project. Contains three short files Claude reads on demand (not every session). The routing map in INSTRUCTIONS.md points to them.
Project memory
Cowork's automatic, project-scoped recall — separate from MEMORY.md. Stored locally on your machine, not in Anthropic's cloud. View, edit, or reset from Project settings > Memory.
Scheduled task
A recurring job set with the /schedule command. Runs at a fixed time, writes results back to your folder. Requires Claude Desktop to be open. Skips and retries when the machine wakes.
Mount mode
The permission posture for a connected folder: Read-only, Read-write, no delete, or Read-write. Set when you connect; change later from project settings.
Ask before acting
The default permission mode. Cowork pauses before each file change and asks you to confirm. Recommended for new users and any session touching files outside output/.
Act without asking
The faster permission mode. Cowork runs through its plan without pausing. Trade-off: less friction, less oversight. Switch to this only for workflows you have run successfully several times already.
Skill
A packaged capability Claude can use — e.g. the trelleborg-cowork-starter skill that scaffolds your folder. Skills run inside Cowork and extend what it can do. They are not the workspace itself. Managed under Customize.
Plugin
A bundle of skills and connectors packaged together for a specific role or use case. Adds capability to Cowork; like skills, plugins are components, not the workspace.
Connector
An OAuth-gated integration with an external service (Gmail, Google Drive, Slack, Notion, etc.). Cowork uses connectors before falling back to browser or screen interaction. Set up under Customize.
Live artifact
A persistent, interactive HTML view pinned to a Cowork project (e.g. a dashboard, a planner). Available since the April 2026 GA release. Local to your machine; not shareable across devices.
Dispatch
Assigning a Cowork task from your phone (Pro and Max plans). The task runs on your desktop machine. A single continuous thread; no multi-thread management.
Computer use
The research-preview feature that lets Claude click, type, and navigate outside the connected folder — your full screen becomes the surface. Not used in the Trelleborg pilot.
Lethal trifecta
Simon Willison's framing of the three conditions that make an AI agent dangerous: reads untrusted input, touches sensitive resources, can act outward. Cowork has all three by design. See the Safety page.
Claude Code
Anthropic's developer command-line tool. Different product from Cowork; lives in a terminal. The Trelleborg starter skill does not use Claude Code — it runs inside Cowork.
claude.ai Projects
Cloud-based conversation folders on the claude.ai website. No local file access, no agentic execution. Unrelated to Cowork despite the name overlap.
§3 — Feedback

Feedback

This guide is a starting point. Every Trelleborg pilot participant is expected to share what worked, what did not, and what was confusingly worded. If a section on this site needed three reads before it made sense, that is a bug — let us know which one.