Business Brain Plugin
🎥 Video lesson — video not yet migrated (source Skool video id:
b3e9657816dc42179183784efbe0e5a0).
Business Brain: how to give Claude a memory of your business
Your AI agents are only as good as the context they have. Right now, every time you start a new task in CoWork, Claude knows nothing about you. This lesson fixes that.
What the brain plugin actually is
It's a folder structure with a specific job: store everything Claude needs to know about your business in one place, and make that information available across every other plugin you build.
Content creation, sales outreach, marketing, operations, it doesn't matter. Any plugin that needs to know your brand guidelines, your company structure, your customers, your competitors, it pulls from the brain. You build it once, and every agent in your stack gets smarter automatically.
Think of it as a persistent memory layer that sits above all your other workflows.
The 3 folders and what they do
The brain plugin uses 3 folders. Each has a specific purpose, and it matters.
Materials — Drop raw source documents in here. Slack exports, meeting transcripts, proposals, financial reports, anything relevant to your business. Run the Ingest Materials capability and Claude will extract the useful information and push it into the right places.
Context — This is where the actual business knowledge lives, as formatted markdown files. There's a CLAUDE.md file in here that acts as a router. Claude reads that first, sees what else is in the folder, and only loads what it needs. This keeps things fast and doesn't bloat the conversation.
Memory — A simple persistent memory system built into the plugin so Claude can retain things across separate conversations.
Setting it up: the order matters
- Upload the plugin zip in CoWork via the Customise tab
- Create a working folder for it (Documents > CoWork Context > Business Brain is a clean structure)
- Run the
0-setupskill first. This builds the folder structure and creates all the template files. Nothing is populated yet, it's just scaffolding. - Then run the full interview from the brain skill menu
Don't skip the working folder step. That's where all outputs get saved. If you're not working from the right folder, nothing will persist correctly.
Running the interview
Once setup is done, open the brain skill. You'll see overall completeness sitting at 0% across 10 domains. That's your starting point.
The fastest way to get meaningful coverage is to give Claude as much to work with as possible before you start talking. This prompt is a solid starting point:
Can we please do the full interview process, and can you use Exa MCP to scrape the web and get as much context about me and my business as possible to help populate the brain. You can also connect to my CRM and pull whatever information is available there.
If you use HubSpot, Slack, or any other connected tool, mention it. Tell Claude to pull from all of them. The goal is to front-load as much context as possible so you're not answering questions you've already got documented somewhere.
One session like this will typically get you to 70-80% completeness. Run it on the latest model for best accuracy. The remaining 20% comes from adding more materials and running domain-specific interviews over time.
Connecting the brain to other plugins
Once the brain is built, it doesn't help anyone sitting in isolation. When you're working in another plugin (say, your content plugin), add the brain folder as an input. The CLAUDE.md in the brain folder tells that agent what information is available and where to find it. So your content agent knows to look there for brand guidelines, tone of voice, audience information, whatever you've stored.
You can also just mention it in the chat: "Check the business brain for our brand positioning before you write this."
The key principle: if you find yourself repeating the same context to multiple plugins, that context belongs in the brain. Move it there once, reference it everywhere.
Visualising what you've built
There's a visualise graph capability in the brain skill that generates an interactive HTML file. Open it in Chrome and you can see how all the knowledge nodes connect, competitors, products, team structure, customers. It's useful for spotting gaps and gives you a clear picture of how much coverage you actually have.
What this gets you
From one solid interview session, most people see an 80% improvement in how accurately their AI agents respond. That last 20% is about specificity, but the big jump happens fast.
The more materials you drop in over time, the more accurate everything downstream becomes. It's not a one-time setup, it's a library you keep adding to.
What to do next
- Run the 0-setup skill and create your working folder
- Drop any existing documents (proposals, transcripts, brand docs) into the materials folder
- Run the full interview with the prompt above
- Visualise the graph and check where the gaps are
- Identify one other plugin you're building and connect the brain to it
Once the brain is running, every other plugin you build gets smarter from day one.