Provision
Describe the infrastructure you need. Get a plan you can read.
You: We need a staging cluster for the orders API in eu-west-1.
SysJ: Will it serve customer traffic, or only your team? And should it be protected from deletion?
You: Team only. No protection needed.
SysJ: A two-node EKS cluster in a private network, reachable through your VPN. I have drafted the plan for you to review.
At a glance
What Provision gives you.
Questions you can answer
The assistant asks whether this is for development or production, how much traffic to expect and what it may cost. It never asks which CNI plugin you prefer. Opinionated defaults cover the rest.
Checked against your account
Regions, machine sizes and connected accounts come from your workspace, so a request can never name something your cloud cannot provide.
Rendered, never written
The model picks modules and fills in parameters. Terraform is generated mechanically from a typed graph, so there is no hand-written HCL to go quietly wrong.
The provider decides what is valid
Every resource is validated against the cloud provider’s own schema before a plan runs. What the assistant can build grows with the provider, not with a catalogue someone maintains by hand.
In detail
How it works.
From a sentence
A conversation that ends in a structure, not a paragraph.
You describe the outcome: a cluster for the API, a Postgres database for orders, a bucket for uploads. SysJ turns that into a resource graph with names, sizes, regions and dependencies in it, and shows you the graph before anything is planned.
- Asks only what it cannot infer
- Explains every default it chose, and lets you change it
- Remembers earlier conversations in your workspace
You: We need a staging cluster for the orders API in eu-west-1.
SysJ: Will it serve customer traffic, or only your team? And should it be protected from deletion?
You: Team only. No protection needed.
SysJ: A two-node EKS cluster in a private network, reachable through your VPN. I have drafted the plan for you to review.
Two editors
Change it in a form when a conversation is the wrong tool.
Nobody wants to have a conversation about renaming a tag. The console edits the same desired state the assistant writes, shows the fields that are safe to change, and produces a plan the same way. Two editors, one path to your cloud.
- Size, count, version and exposure as plain fields
- The generated Terraform is always one click away
- Both editors end at the same plan and the same approval
Every major cloud
One loop across AWS, Google Cloud and Azure.
The same describe, plan, approve and apply loop runs against each provider, with delegated access on each: an IAM role on AWS, Workload Identity Federation on Google Cloud, and federated credentials on Azure. SysJ never holds a long-lived key.
- Amazon EKS, Google GKE and Azure AKS clusters
- Networking, databases, storage and identity
- Terraform state kept in a bucket in your own account
Questions
Common questions about Provision.
- Does the AI write Terraform?
- No. The model chooses modules and parameters inside a typed graph. Terraform is rendered from that graph by code, validated against the provider’s schema, and planned for real before anyone is asked to approve it.
- What if my request is ambiguous?
- The assistant asks. It asks in terms a product engineer can answer, such as whether this serves production traffic, and it states the default it will use if you do not mind.
- Can I see the Terraform?
- Always. It is committed to a repository you own, and the console shows it beside the plain-language summary.
Keep reading
The rest of the platform.
See your own infrastructure as a plan.
Tell us about your team and the cloud you use. We will set up your workspace and walk through your first plan with you.