BRG AI Enablement  ·  August 2026  ·  v5

Build Your Own
CoworkOS

Give Claude persistent memory, your voice, and custom workflows — in one afternoon.

Cowork  ·  Claude Desktop App ·  Brought to you by Vivid
What a workstation actually is — and how it differs from a project Why CoworkOS matters and the three core files behind it Starter templates — copy and paste to get going Prompts for your first scheduled task and your first workstation
BRG AI Enablement
BRG

Claude Tools — When to Use What

A quick reference for BRG business users across all teams

Browser  /  Mobile  /  Desktop
Claude Chat
claude.ai  ·  No setup required
Reach for it when
You need a smart thinking partner — right now, from anywhere.
Good for
Quick research & summarisation
Paste a doc, ask a question, get a clear answer fast
Drafting & editing on the fly
Emails, Slack messages, short reports, meeting summaries
Brainstorming & thinking out loud
Workshop prep, idea generation, problem framing
One-off file analysis
Upload a PDF, image, or CSV and ask questions about it
Not ideal for
Accessing files on your computer, automating recurring tasks, or work that needs memory across sessions
Desktop App (Mac  /  Windows)
Claude Cowork
Installed locally  ·  Connects your tools
Reach for it when
You want AI to actually do work inside your tools and files — not just talk about it.
Good for
Multi-step workflows across tools
Read Slack, draft a reply, update a doc, send an email — in one go
Working directly with local files
Build decks, spreadsheets, Word docs from your own folders
Scheduled & repeating tasks
Morning digest, weekly reports — set once and forget
Persistent memory & workstations
Claude remembers your context, tone, and rules across sessions
Not ideal for
Writing or modifying code in a codebase, running tests, or engineering workflows needing terminal access
Terminal  /  Command Line (CLI)
Claude Code
Developer tool  ·  Requires setup
Reach for it when
You're building, fixing, or shipping software — AI works alongside your actual codebase.
Good for
Writing & debugging code
Features, bug fixes, refactors across real project files
Scripting & automation engineering
Python scripts, data pipelines, CLI tools, APIs
Code review & testing
Run tests, review PRs, catch issues, explain unfamiliar code
Agent & MCP development
Build custom AI tools, integrations, and Cowork plugins
Not ideal for
Non-technical users, everyday comms tasks, or anything that doesn't involve writing or running actual code
The Problem

Without a system, Cowork forgets you. CoworkOS fixes that.

Without Cowork OS

A fresh start every time

Every session begins from zero. Claude doesn't know your name, your role, your preferences, or what you were working on yesterday. You spend the first few minutes re-explaining context before any real work happens.

Generic tone. No memory. No shortcuts. Repeat yourself constantly.

CoworkOS is just a folder on your computer. No external services, no lock-in.

The One Idea

A workstation is a specialist you brief once.

You explain the job properly one time — the files, the rules, the format, the people involved. From then on you just ask for the thing, and it already knows. Like a colleague who has done your bank rec forty times and doesn't need the background again.

Does this job deserve its own workstation? Three questions
Does it come back?
Weekly, monthly, every close. Not once. If you'll only ever do it again in theory, it's not a workstation.
Do you re-explain yourself?
Same files, same rules, same "no, not like that" correction. If you're typing the same context each time, that's the brief you should be saving.
Does "done" look the same each time?
A shape you'd recognise — same columns, same format, same checks. If every output is different, keep it in a normal chat.
Three yeses
Build a workstation. You'll get the hour back within a month.
Any no
Just ask Claude normally. No folder, no setup, no guilt.

Most of your work is not a workstation, and that's the point. They're for the five jobs you've already done a dozen times and will do a hundred more.

The Question Everyone Asks
READ THIS ONE TWICE

The project is the context. The workstation is the do-er.

Four things that sound similar and aren't. Nine out of ten mix up the first two.

Claude Project
The context
In claude.ai  ·  a shared bucket
Think of it as
A folder of background reading you hand Claude before a conversation.
It holds
Reference documents
Policies, past reports, source files
Standing instructions
Custom instructions for that topic
A chat history
Conversations grouped by subject
It does not
Touch your local files, follow a workflow, or run on a schedule. It waits to be asked.
Workstation
The do-er
In CoworkOS  ·  a folder that acts
Think of it as
A specialist you brief once, who then does the job on request.
It holds
A workflow
The steps, in order, every time
Its own rules and memory
What good looks like in this job
A place to put the output
Saved where you expect to find it
It does not
Suit one-off work. If the job won't come back, you're doing setup for nothing.
Skill
The tool
Installed once  ·  used anywhere
Think of it as
A power tool on the bench. Any workstation can pick it up.
It gives Claude
One repeatable capability
Build a spreadsheet, write a Word doc
Known-good method
Same technique whoever calls it
Shareable packaging
Hand it to a colleague as one file
It does not
Know your job, your files, or your standards. It's a capability, not a colleague.
MEMORY.md
The notebook
One per workstation  ·  plus a root one
Think of it as
What the specialist wrote down so you'd never have to say it twice.
It holds
Facts that change
Who owns what, which entity, which path
Decisions and why
So last month's call still stands
Corrections you've made
Say it once, it sticks
It does not
Hold rules. "Always do X" belongs in CLAUDE.md. Memory is facts, not instructions.
Already built Claude Projects? Nothing is wasted. The reference files in your project become the workstation's Resources folder, and your custom instructions become its CLAUDE.md. Ask Cowork: "I have a Claude Project for [topic] — help me turn it into a workstation and tell me what should stay a project." You're moving context into a place that can act on it, not starting again.
A Worked Example

Same Monday, twice. One version takes four minutes.

Chat, or a project

You do the explaining, every week

  1. Find last week's file. Upload it.
  2. Explain which columns matter and which to ignore.
  3. Re-explain the tolerance rule you mentioned last time.
  4. Get output in the wrong format. Correct it.
  5. Copy the result somewhere you'll remember.
40 minand it resets next Monday

Not hypothetical. An AR specialist in APAC built three reconciliation workstations that now run in live monthly use. A finance manager in Canada built five.

How It Works

Three layers that power CoworkOS. One folder to rule them all.

You  —  prompts, questions, and tasks
CoworkOS
CLAUDE.md + MEMORY.md
Global rules, preferences, routing map, and accumulated facts
voice-principles.md + Skills
Your writing voice and installed capabilities
Workstations
Bank Rec Month-End Close Email HQ Weekly Report Slide Decks + yours

Each workstation is one specialist, in one folder: a local CLAUDE.md for its rules, a MEMORY.md for what it has learned, and a Resources subfolder for its reference files. The global files apply everywhere; the workstation files layer on top.

The Core Files

Three files power the whole system

File 01
CLAUDE.md

The global constitution. Tells Claude your preferences, rules, folder structure, and routing map. Every session starts by reading this.

File 02
MEMORY.md

The persistent brain. Claude reads it at session start and writes to it when you say "remember this." It builds up your context over time.

File 03
voice-principles.md

Your writing voice. Tone, sentence style, words to avoid, signature phrases. Claude reads this before writing any content on your behalf.

These three files live in your CoworkOS root folder alongside a 00_Resources/ folder. That's the whole system.
Setup Guide  ·  Three Phases

Let's get started.

You're one afternoon away from a Claude that knows who you are, how you work, and what you're trying to get done.

📁
Phase 01
Create your home base
Make an empty CoworkOS folder on your computer and run the setup prompt — Claude builds the whole structure for you.
🏗️
Phase 02
Get one win, then build
Start with one scheduled task that runs itself. Then turn your most repeated job into a workstation — its own rules, its own memory.
🔄
Phase 03
Keep it compounding
Save insights at the end of every session. Claude gets sharper over time — not because it learned, but because you built the system.
Phase 1 | Getting Started  ·  Prompt Template

Paste this to build your CoworkOS from scratch

First-time setup

One prompt. Five minutes. 1st part Done.

Open Cowork with your CoworkOS folder selected, paste this prompt, and let it run. Claude will create all three core files, the folder structure, and then ask you a few questions to personalise everything.

Before you paste: create an empty folder called CoworkOS somewhere on your computer (Documents is fine), then select it in the Cowork file picker.
I'm setting up CoworkOS for the first time. Please
create the following in my CoworkOS root folder:

1. CLAUDE.md — with these sections:
   Memory System (include memory routing rules:
   root vs workstation MEMORY.md, and that
   behavioral rules go to CLAUDE.md not MEMORY.md),
   Preferences, Rules, Folder Structure, Routing Map

2. MEMORY.md — with these sections:
   About Me, Active Projects, Key Contacts,
   Key Decisions, Useful Context

3. 00_Resources/voice-principles.md — with these
   sections: Tone, Sentence Style, Words to Avoid,
   Words I Use, Formatting

4. 00_Skills/        (empty folder)
5. 00_Scheduled/     (empty folder)

Use starter templates with [bracketed placeholders]
where I need to customise. After creating the files,
ask me 3 questions to personalise my CLAUDE.md for
my role and working style.
After Running the Setup Prompt

Let's see what you just created.

That one prompt created your entire CoworkOS skeleton — three files, two empty folders, and a structure Claude will use on every future session.

The next three slides walk through each file with a copy-ready template. Customise the bits in [brackets] and you're good to go.

Once you've filled in the templates, add your first workstation. That's when things start to feel like a real operating system.
📁 CoworkOS/
├── 📄 CLAUDE.md  — global rules & routing
├── 📄 MEMORY.md  — persistent brain
├── 📁 00_Resources/
    └── 📄 voice-principles.md
├── 📁 00_Skills/  empty for now
└── 📁 00_Scheduled/  empty for now
After adding a workstation:
└── 📁 Email HQ/
    ├── 📄 CLAUDE.md
    ├── 📄 MEMORY.md
    └── 📁 Email HQ Resources/
File 01  ·  Starter Template

CLAUDE.md — the global constitution

Copy & customise

Sets the rules for every session

This file is Claude's operating instructions. It loads automatically and governs tone, behaviour, folder structure, and which workstation to use for which task.

Replace anything in [brackets] with your own preferences. The routing map grows as you add workstations.
# CLAUDE.md

## Memory System
Read MEMORY.md at the start of every session.
Write to MEMORY.md when I say "remember this."

+ new
Memory routing — which MEMORY.md?
- Cross-cutting (role, people, preferences, projects)
  → root MEMORY.md
- Workstation-specific (domain contacts, decisions,
  output paths, tool configs) → [Workstation]/MEMORY.md
- Behavioral rules (always/never/before doing X)
  → CLAUDE.md (root or workstation), not MEMORY.md
Never duplicate the same fact across both files.


## Preferences
- Tone: [conversational / professional / casual]
- Language: [Australian / British / American English]
- Length: concise by default; expand only when I ask
- Format: prose first, bullets only when content is
  genuinely list-like

## Rules
- Ask clarifying questions before complex tasks
- Never guess — if unsure, say so
- Read voice-principles.md before writing any content
  on my behalf

## Folder Structure
| Folder              | What lives here                |
| 00_Resources/       | Reference files (voice, etc.)  |
| 00_Skills/          | Installed skill archives       |
| 00_Scheduled/       | Scheduled task outputs         |
| [Workstation]/      | Workstation + Outputs/ folder  |

## Routing Map
| Workstation         | Route here when I...           |
| [Workstation Name]  | [Brief trigger description]    |
File 02  ·  Starter Template

MEMORY.md — the persistent brain

Copy & customise

Claude reads and writes this automatically

Fill in the About Me section yourself — the rest Claude populates over time as you work together. Don't over-engineer this on day one; let it grow.

The Key Decisions section is intentionally left empty. Claude adds to it whenever it makes a non-obvious choice — handy for future sessions.
# MEMORY.md

## About Me
- Name: [Your name]
- Role: [Job title, team, company]
- Key priorities: [Your top 2–3 focus areas]
- Working style: [e.g. async-first, direct comms]

## Active Projects
| Project  | Status | Notes               |
| [Name]   | Active | [One sentence]      |

## Key Contacts
| Name     | Role       | Why they matter    |
| [Name]   | [Role]     | [Context]          |

## Key Decisions
(Claude populates this over time as we work —
don't edit manually)

## Useful Context
(Claude adds facts here when you say
"remember this")
File 02  ·  Memory Routing

Two notebooks. One question tells you which.

Ask yourself: would this be useful in three different jobs, or only this one? Three jobs, it goes in the root notebook. Only this one, it goes in the workstation's own. That's the whole rule.

Root MEMORY.md

Things true no matter what you're working on — your role, the people you deal with, what you're currently on, how you like things done. Useful in three different jobs? It belongs here.

CLAUDE.md (root or workstation)

Rules, not facts. Anything with an "always", "never" or "before you do X" in it. If you're telling Claude how to behave rather than what's true, it goes in CLAUDE.md instead.

The golden rule

When in doubt, write to root MEMORY.md. Never duplicate the same fact across both files. If a memory could plausibly live in either place, root wins — it's always in scope regardless of which workstation you're in.

File 03  ·  Starter Template

voice-principles.md — your writing voice

Copy & customise

Makes Claude sound like you, not a press release

Paste 5–8 of your own emails or Slack messages and ask Claude to extract patterns from them. It'll fill in the signature phrases and tone sections automatically.

The more specific you are about what to avoid, the better. "No em dashes" and "never say leverage" are more useful than "be casual."
# Voice Principles

## Tone
- Sound like: [smart peer / trusted advisor / warm colleague]
- More formal with: [senior leaders / new contacts]
- Avoid sounding: [corporate / over-hedged / AI-flavoured]

## Sentence Style
- Vary length — mix short punchy beats with longer ones
- Use contractions: [yes / no]
- Front-load context, then the ask

## Words to Avoid
- Never: "leverage," "synergise," "deep dive," "circle back"
- No em dashes (use commas or parentheses instead)
- No "I hope this email finds you well" openers
- No stacked adjectives ("incredibly valuable", "seamless")

## Words I Use
- [Your signature phrases — add them here]
- e.g. "heaps", "nerd out", "level up", "if helpful"

## Formatting
- Short paragraphs (1–2 sentences max)
- Bold sparingly — key terms and names only
- [Other formatting preferences]
Phase 02  ·  Get Something Running

Win first. Build second.

Don't start with a workstation. Start with one scheduled task that saves you a job you resent — it takes two minutes and it proves the whole idea. Build the workstation once you already believe it.

Phase 2  ·  Start Here

The two-minute win: one job that runs itself

Before your first workstation

Belief is cheaper to buy than to argue for

A scheduled task is Cowork doing one small job on a timer, without being asked. No folders, no files, no setup. It is the fastest way to see the point.

Pick the check you do every morning out of habit — a balance, a queue, an inbox, a report that lands late. Tell Cowork to do it and message you the answer.

A credit manager who rated her own confidence 2 out of 5 shipped two of these in her first fortnight. One compares daily cash to forecast and posts the gap to Slack. Confidence followed the win, not the other way round.
Set up a scheduled task for me.

What I check manually right now:
[describe it plainly — e.g. "every morning I open
the bank file and compare actual cash to the
forecast, then flag anything more than $5k out"]

When it should run:
[e.g. "every weekday at 7am Sydney time"]

How I want the answer:
[e.g. "a short Slack message to me — just the
variance and anything unusual, not a full report"]

Before you set it up:
- Tell me what you'll need access to
- Run it once now so I can see the output
- Then confirm the schedule with me

If anything is ambiguous, ask me rather than
guessing.
Phase 2 | Workstations

Four things to tell your new specialist

This is the brief from slide 4, written down. Four short sections, in one file, so you never have to say any of it again.

Resources

A table mapping reference files to "read when" conditions. Claude only loads these files when the trigger is met, keeping context efficient.

Workflow

Numbered steps for the primary task in this workstation. Keep it simple — 5–7 steps. You'll refine it over time as you see what Claude gets right and wrong.

Editorial Rules

Always opens with "Follow voice-principles.md in 00_Resources." Then adds domain-specific rules on top — tone calibration, vocabulary, output formats for this specific domain.

Phase 2 | Workstations  ·  Prompt Template

Paste this to create a new workstation

New workstation

30 seconds to a new specialist

Pick a job that passed all three questions on slide 4 — a reconciliation, a weekly report, a recurring email. Fill in the two brackets, paste, and Claude scaffolds the whole thing.

Good workstation triggers are specific: "draft, edit, or review any Slack message" rather than "Slack stuff." The routing map entry is what Claude matches against your request.
Create a new workstation called "[Workstation Name]".

Route here when I: [describe the task type — e.g.
"draft, edit, or review any Slack message"]

Create these three items inside the workstation folder:

1. [Name]/CLAUDE.md with sections:
   - Identity (one paragraph: what routes here,
     what doesn't)
   - Resources (empty table: Resource | Read when...)
   - Workflow (numbered steps for the main task)
   - Editorial Rules (start with voice-principles.md
     reference, then add domain-specific rules)

2. [Name]/MEMORY.md with:
   - Contacts section (people relevant to this domain)
   - Key Decisions section (domain-specific choices)
   Note: cross-cutting facts (role, people, projects)
   belong in root MEMORY.md, not here.

3. [Name]/[Name] Resources/  (empty folder)

After creating the files, add a new row to the
Routing Map in root CLAUDE.md.
A Quick Recap

One folder. A few global rules. As many specialists as you need.

You  —  prompts, questions, and tasks
CoworkOS
CLAUDE.md + MEMORY.md
Global rules, preferences, routing map, and accumulated facts
voice-principles.md + Skills
Your writing voice and installed capabilities
Workstations
Bank Rec Month-End Close Email HQ Weekly Report Slide Decks + yours

If you remember one thing: a workstation is a specialist you brief once. The rest — folders, files, routing — is just how that brief gets stored so Claude can find it again next week.

Phase 03  ·  System Maintenance

Last bit. Make it compound.

The system only gets smarter if you save what you learn. One 30-second prompt at the end of each session — that's all it takes to turn CoworkOS from a config file into something that actually grows with you.

Phase 3 | System Maintenance

CoworkOS only gets smarter if you save what you learn

End of every task

Insights vanish when the session closes

Preferences you mentioned, decisions you made, context Claude picked up on — none of it survives unless it's written to MEMORY.md. The AI resets completely.

One prompt at the end of every task fixes this permanently. Paste it into Cowork and Claude scans the whole conversation, then writes anything useful to the right place.

Done consistently, this is what turns CoworkOS into a system that compounds — every session makes the next one sharper.
💬 Work
session
🔍 Scan
convo
📝 Write to
MEMORY.md
Smarter
next time
Scan this conversation for any unsaved preferences,
decisions, or useful context. For each item found,
route it to the right place:

- Cross-cutting (role, people, projects, preferences)
  → root MEMORY.md
- Workstation-specific (contacts, decisions, paths,
  tool configs for this domain)
  → [Workstation]/MEMORY.md
- Behavioral rule (always/never/before doing X)
  → CLAUDE.md (root or workstation)

Don't duplicate what's already there. Focus only
on what's genuinely new from this session.
Scale your workstation

Pick the one job you resent repeating. That's your first workstation.

CoworkOS compounds. The more you use it, the more Claude knows you. Start small — one folder, three files — and add workstations as you go.

Questions? Find us in #bai-claude-ama on Slack.

01
Create your CoworkOS folder
Make an empty folder in Documents (or wherever). Select it in the Cowork file picker. That's your workspace.
02
Paste the first-time setup prompt
Slide 10 has the prompt — hit Copy, open Cowork, paste, let it run. Three files and a folder structure in under a minute.
03
Get one win, then build one workstation
Ship the scheduled task on slide 17 this week. Then use the workstation prompt on slide 19 for the job you repeat most.
04
Share what you build
Drop a note in #bai-claude-ama — what workstation you built, what it does, what surprised you. Help the next person skip the setup.
Efficiency

Keep your CoworkOS fast and cheap

Rule 1
Constrain the Root

Keep your master CLAUDE.md under 300 lines. Point to external resource files rather than stuffing everything in one place — this keeps baseline token costs low every single session.

Rule 2
Avoid Duplicated Rules

Never repeat global rules inside local workstations. If the root already says "no corporate jargon," your Email HQ workstation shouldn't waste tokens restating it. Inheritance is free.

Token efficiency means CoworkOS stays fast. A bloated root file taxes every session — trim it like you'd trim a good document.
Reference  ·  Choosing a Model

Which Claude model, and when

Four models, one dial: bigger = smarter and pricier. Match the model to the stakes, not the excitement.

Fastest & cheapest
Haiku 4.5
Reach for it when
The job is small, well-defined, and high-volume.
Good for
Classification & data extraction
Quick lookups & short summaries
Routing & triage steps in an agent
Simple transforms & translation
Lowest cost. Its sweet spot is doing one simple thing thousands of times.
The everyday workhorse
Sonnet 5
Default
Reach for it when
Almost always. It covers ~80% of daily CoworkOS work.
Good for
Drafting emails, docs & decks
Research & synthesis
Most coding & code review
Everyday chat & workflows
Fast & well-priced. Start here — only reach higher when it visibly struggles.
The heavy thinker
Opus 5
Reach for it when
The task is genuinely complex — 3+ interdependent steps.
Good for
Multi-entity consolidations
Board-level or high-stakes writing
Greenfield architecture & gnarly debugging
Planner role in agentic work
$5 / $25 per M. Near-frontier quality at half Fable's price — worth it when the stakes are high.
Mythos-class · above Opus
Fable 5
Reach for it when
Long-horizon autonomous runs — its lead grows with task length.
Good for
Managed Agents running unattended
Deep long-context, file-based memory work
State-of-the-art finance & vision tasks
Not everyday human-in-the-loop work
$10 / $50 per M · 30-day data retention. Overkill for daily work — weigh retention for sensitive data.
Reference

Glossary of Terms

Model The underlying AI engine. Claude's current lineup: Haiku 4.5 is fastest and cheapest for lightweight jobs; Sonnet 5 covers ~80% of daily work; Opus 5 is for complex, multi-step tasks with many interdependent decisions; Fable 5 is the Mythos-class model above Opus, reserved for long-horizon autonomous agent runs. ⚡ Default to Sonnet 5. Reaching for Opus (or Fable) reflexively just burns tokens — and slows you down. Most tasks don't need it. Don't be a #TokkenMaxxer. Think of the environment.
Agent An AI model that can take actions — read files, send messages, run tasks — rather than just answer questions. Claude in Cowork mode is an agent.
Workstation A specialist you brief once. A CoworkOS folder for one recurring job, holding its own rules, memory and resources. Claude routes work to the right one using your routing map.
Project (in claude.ai) A bucket of reference files and standing instructions for one topic, in the Claude web app. The project is the context; a workstation is the do-er. A project waits to be asked; a workstation runs a workflow on your local files.
Skill A packaged workflow that gives Claude a specialised capability — creating decks, writing Word docs, pulling data. Install once, trigger by name.
Session A single conversation with Claude. Context resets when it ends — which is exactly the problem MEMORY.md solves.
MCP Model Context Protocol. The open standard that lets Claude connect to external tools and services — Slack, Jira, email, and so on. Think of it as Claude's plugin API.
Context Window The total amount of information Claude can hold in a single session — your instructions, files, and conversation history. Bigger context = more Claude knows, but also slower and pricier.
MCP Server A connector that implements MCP for one specific service. The Slack MCP server is what gives Claude read/write access to your Slack workspace.
Token The unit any LLM uses to measure text in and out. Longer prompts cost more tokens. The reason Sonnet or Haiku is faster and cheaper than Opus for everyday tasks.
Plugin A bundle of MCP servers and skills packaged together for easy installation. How capabilities are distributed and added to Cowork.
Routine A task set to run automatically on a schedule. "Every morning, summarise my unread Slack." Cowork runs these in the background without you having to ask each time.
Thank you
Questions? Find Vivid or Lilly on Slack.
#bai-claude-ama
1 / 23
← → to navigate