Building a personal Claude operating system
Wire Projects, Memory, connectors, and a scheduled task to your week
After this, you'll be able to combine the features from this whole track into a single configured setup that makes Claude your default tool for recurring work, with one concrete worked example to copy.
Before you start
Complete Sharing Claude with your team: what needs setup and what doesn't first; this lesson builds on that account-level versus plan-level distinction, because your Personal Claude OS lives on your individual account and you need to know what stays yours.
The idea
A Personal Claude OS wires Projects, Memory, connectors, and a scheduled task together into your default tool for recurring work. On their own, each feature is useful. Wired around how you actually work, they stop being features you visit and become the place your recurring work lives.
The point is not to use every feature. It is to wire the few that fit your week into one setup so you reach for Claude by reflex instead of starting from a blank chat each time. One tier note: Projects, Memory, and connectors are free, while a Cowork scheduled task is a paid Desktop feature, so a free learner can build the first three layers today and add the scheduled task on Pro.
Here is the before and after: Without a setup, a consultant opens a blank chat every morning, re-explains her clients, and re-pastes her style guide. With a Personal Claude OS, a "Client Work" Project holds her role and each client's brief, Memory carries her standing preferences, a Google Drive connector reads the real files, and a Cowork scheduled task on Desktop delivers her Monday digest before she sits down.
Now try it sketch your own layers on paper: one Project, the two or three standing preferences you want Memory to keep, one connector, and (if you have Pro) one Cowork scheduled task, each tied to a task you actually repeat.
A Personal Claude OS is a few features wired to your real week, so recurring work starts already set up instead of from scratch.
Try it (18 min)
Watch out for
- Trying to wire every feature at once; build one layer, confirm it works, then add the next, or you will not know which piece broke
- Putting standing preferences in a Project's Instructions when they belong in Memory (or the reverse), so they apply too narrowly or too broadly
- Expecting the Cowork scheduled task on free; it is a paid Desktop feature, so plan the first three free layers and add it once you are on Pro
- Adding a read+write connector without noticing it can change your real data, then being surprised when a scheduled task acts on it
- Building the setup once and never revisiting it, so it drifts out of sync with how your week actually changed
Paste this into Claude
I want to build a Personal Claude OS (a setup of Projects, Memory, connectors, and a Cowork scheduled task on Desktop wired to my actual recurring work) so I stop starting from a blank chat every day. Note on tiers: Projects, Memory, and connectors work on free; a Cowork scheduled task is a paid Desktop feature. Here is how I actually work: MY RECURRING WORK: - [e.g. "I write weekly client updates for 3 clients"] - [e.g. "I prep for the same Monday team meeting every week"] - [e.g. "I keep all my reference docs in Google Drive"] MY STANDING PREFERENCES (things I'd tell any assistant once): - [e.g. "I write in plain English, short paragraphs, no jargon"] - [e.g. "I'm in the Pacific time zone"] Please design my Personal Claude OS as a concrete setup with four layers, marking which layers I can build on free and which need Pro: 1. PROJECTS: which Project(s) to create, and what the Instructions for each should say (free) 2. MEMORY: which standing preferences I should set through Memory's custom instructions so they carry across every conversation (free) 3. CONNECTORS: which connector(s) fit my work, and whether each is read-only or read+write (free) 4. SCHEDULED TASK: at least one Cowork scheduled task on Desktop to build, what it pulls from and where it delivers, and a note that this layer needs a paid plan Then give me a build order: which layer to set up first, second, third, and fourth, and why that order.
What good looks like
- Claude designed all four layers (Projects, Memory, Connectors, a Cowork scheduled task) tied to your real recurring work, not generic examples
- Claude marked the first three layers as free and the Cowork scheduled task as a paid Desktop feature
- The Project layer includes actual Instruction text you could paste, not just a description
- Each connector is labeled read-only or read+write so you know what Claude can do with it
- Claude gave you an ordered build sequence with a reason for the order
When this breaks
- Breaks when a scheduled task depends on a connector or Project you have not set up yet because it fires on an empty foundation and delivers a blank or wrong result while you assume it is working.
- Degrades when you copy a generic 'ultimate setup' instead of wiring features to your real recurring tasks because you maintain layers you never use and the setup feels like overhead instead of real help.
AI can help with this
Paste this: 'Design my Personal Claude OS. My recurring work: [list]. My standing preferences: [list]. Give me four layers, marking which are free and which need Pro: 1) Projects with paste-ready Instructions (free); 2) Memory facts (free); 3) connectors labeled read-only or read+write (free); 4) a Cowork scheduled task on Desktop with what it pulls and where it delivers (paid). Then give me a build order and why.'
You can now
You can complete the lesson outcome in a real Claude chat, Project, Artifact, Connector, Desktop, or Code surface.
- You can verify that claude designed all four layers (Projects, Memory, Connectors, a Cowork scheduled task) tied to your real recurring work, not generic examples.
- You can verify that claude marked the first three layers as free and the Cowork scheduled task as a paid Desktop feature.
- You can verify that the Project layer includes actual Instruction text you could paste, not just a description.
- You can verify that each connector is labeled read-only or read+write so you know what Claude can do with it.
Key takeaways
A Personal Claude OS wires Projects, Memory, connectors, and a scheduled task to your real recurring work, so the stage is already set when you start. The first three layers are free; the Cowork scheduled task is a paid Desktop feature.
- Combine Projects, Memory, connectors, and a scheduled task around your real week, not around using every feature.
- Put standing preferences in Memory and role plus context in a Project's Instructions, so each applies at the right scope.
- Build foundations first: a scheduled task needs its connector and Project in place before it can deliver real output.
- Three layers (Projects, Memory, connectors) are free; the Cowork scheduled task needs a paid plan, so build what you can today and add it on Pro.