Agency AI Infrastructure Blueprint
Build the AI foundation
for your team’s client work.
Build a setup your team can use to turn client drafts into work that is internally checked, corrected and ready to send. The Blueprint gives you the templates, walkthroughs and build order. Your team implements the system and does the client work.
- Build it in your own accounts
- Templates + walkthroughs
- Self-tests at each step
Net, one-time. Access by email. VAT shown before payment. All sales are final.
When a colleague needs
the setup you built.
Your AI sessions start with client context you have checked. Then a colleague picks up the next draft for the same account. They need that context and the instructions for preparing the work.
You could send the files. But what happens when the client changes their offer? Which copy will your colleague use next week? And should everyone have access to that client's material?
Until those decisions reach the person doing the work, a review can turn into another search for an offer, a briefing or a correction someone already made. The information needs a way to reach the right people and stay current.
The Blueprint is the kit for building that shared infrastructure. It starts with the source of your context and works outward to access, reusable task instructions, connections and ownership.
Build the environment
around your shared context.
Each part of the Blueprint gives you something concrete: a template to copy, a walkthrough to follow, a decision to record or a self-test to run.
First, give context a shared home.
Put your BrandOS and client contexts in repositories with a clear current version. Set up loading for approved sessions. Write the update routine: change, check, merge, next session.
Then build the environment around it.
Decide who gets access, which connections are approved and where shared Skills live. Name an owner and write down how another person joins the setup. Your team checks each layer before building the next one.
Two systems, in this order
- One shared home for brand and client context
- Loading into every approved session
- Update logic: a change lands once
- The rule that says what counts as current
- The shared workspace and its rules: where things live, what gets reviewed
- Roles and access
- Connections through one approval point
- A shared Skills library
- Governance: owners, update, review. The last step.
Each layer stands on the one before it. Access rules over context that still lives in private copies are paperwork.
The templates help you build the structure. Your team supplies the client knowledge, makes the access decisions and checks the results. The system distributes what you record; deciding whether it is right stays with you.
A practical build order
with a check at every step.
Seven components, grouped into three phases. Each component comes with a short video and written guidance. A separate video takes you through the build order.
Give context a shared home.
Set a current source for each brand, load it in approved sessions, and check that a change reaches the next person.
See the details
1. The shared home. Brand- and client-repository templates, with a README describing the current source and the review rule.
2. Context loading. A walkthrough for Claude Code and Cowork. The check: change a fact at the source and see it in a colleague’s next session.
3. The update routine. The path from a proposed change to the version your approved sessions use.
Set access and connections.
Decide who can use each client’s material and which tool connections the agency approves.
See the details
4. Roles and access. A decision sheet and a walkthrough for GitHub teams and repository rights. The access decision has a default and its trade-off beside it.
5. Controlled connections. The portal template for your Cloudflare account, a deployment walkthrough and a connection approval sheet. You check that an approved connection passes through the portal.
Maintain the work as a team.
Give recurring instructions an owner, with review rules and a clear route for the next colleague to join the setup.
See the details
6. Shared Skills and Plugins. A Plugin scaffold with a Skill from my own system, and the routine for turning repeated task instructions into a maintained Skill.
7. Governance and onboarding. Owner fields, update and review rules, and the checklist your team follows when someone joins.
The build checklist includes a self-test for every step. Start with the brand-repository template and a usable BrandOS. Check the loading on your own setup before adding the next person.
Net, one-time. Access by email. VAT shown before payment. All sales are final.
The important decisions
are still yours.
Who should be able to read a client's context? Who approves the next connection to your project tool? Who keeps an offer current when it changes?
A template cannot answer those questions for your agency. The decision sheets put them in front of you, with a suggested starting point and the trade-offs. Your system owner makes the call and checks the implementation.
This is real technical work in your own GitHub and Cloudflare accounts. The walkthroughs are written for Claude Code and Cowork. You need someone on your side who will do the work and own the result.
The kit does not include implementation or build support from me. That boundary matters: buy it if your team wants to build, and use the self-tests to check what you have built.
The kit for building
your agency’s foundation.
€99net · one-time
Both parts of the infrastructure, with the materials and checks to build them in your own accounts.
- Brand and client repository templates
- Loading and update walkthroughs for Claude Code and Cowork
- Roles, access and connection decision sheets
- Cloudflare portal template and deployment walkthrough
- Skill and Plugin scaffold, governance and onboarding templates
- A video per component, plus the build order
- A checklist with a self-test at every step
Access to the files and videos arrives by email through your login area on csauerborn.com. Bring a usable BrandOS and someone who will build and maintain the setup. Technical implementation and ongoing support are yours.
Net, one-time. Access by email. VAT shown before payment. All sales are final.
Before you begin.
Do I need to own the Builder or Auto-Load?
No. You need a usable BrandOS before starting the infrastructure build, whether you made it with the Builder or another method. Auto-Load is optional.
Can a non-technical team use it?
The templates and walkthroughs give you a starting point and a build order. Someone still needs to carry out real technical work in GitHub and Cloudflare, make the access decisions and run the checks. This is a kit for doing that work, rather than a managed implementation.
Will it work with other AI environments?
The loading, Skills and connection walkthroughs are written for Claude Code and Cowork. The repository and context structure can carry over to other environments, but you would adapt their loading and connections yourself. Those walkthroughs are not included.
Does it include client-context extraction or support?
Your team prepares the client contexts before putting them in the shared structure. There is no extraction method, build review, support channel or working session included in the Blueprint.
How is this different from the Install?
With the Blueprint, your team builds from the kit. With the Install, I build the agreed infrastructure and hand it over into your accounts. The Install includes the same Blueprint as its manual. The purchases are independent; buying the Blueprint first does not reduce the Install price.
Is this a subscription? Can I get a refund?
It is a one-time purchase. All sales are final. VAT is shown at checkout before payment.
Build a setup
the next person can use.
Put the current client context and working instructions within reach of the next colleague doing the job. Your team builds the setup and takes the work through its final internal review.
Net, one-time. Access by email. VAT shown before payment. All sales are final.