Make every client install repeatable
The steps you repeat at every client should live in a shared, checked process, not in one engineer’s head.
A good client install leaves behind a way to do the next one, not just a machine that works.
Start with one problem
A client install starts with four things: the problem to fix, the systems it runs on, the people responsible, and what proof will count as done. If you pitch everything at once, it’s harder to see the one problem the client is paying you to fix first. The same mini apps serve one person running their own agents and consultants installing them for clients. For each client, write down what’s installed, what support comes with it and what it costs.
Write the install down, step by step
Remote Setup handles the install itself: whether we can reach the server, the plan, trial checks, the steps that were approved, whether it worked, and a record of all of it. Rules & Approvals decides the steps that need permission. Operator Inbox keeps the decisions a person made. Marketplace looks after the client’s connected services and their keys. Other mini apps join when the client needs them, such as Knowledge for documents with sources or Task Board for a shared record of the work. Each one stays its own mini app.
Hand over proof, not just the keys
A repeatable install records:
- the code you started from;
- what was built and installed;
- whether it’s healthy;
- which real job ran;
- whether a person signed it off.
For companies that get audited, some things are not finishing touches: backups, a way to roll back to the last good version, keeping each client’s data separate, and a way to cut off an agent’s access. Turning those steps into mini apps means less reinventing for the delivery team and a clearer system for the client to own after you leave.
