Bicep

Bicep is a full peer language to Terraform on the same canvas. Open your .bicep and .bicepparam files and you get the same graph: params, vars and outputs as nodes, resources as nodes, references as wires, and the value each row resolves to shown in place. The engine is a hand-rolled Bicep lexer/parser/evaluator compiled to the same WebAssembly bundle — no Bicep CLI is required, and your code never leaves your machine.

The Azure quickstart private-endpoint-sql canvas: main.bicep vars flowing into a SQL server, database, VNet, and private endpoint.
The Azure quickstart private-endpoint-sql canvas: main.bicep vars flowing into a SQL server, database, VNet, and private endpoint.

The same canvas, the same moves

Everything you do with Terraform you do with Bicep:

  • Trace — click a row and its whole lineage lights up while everything else dims.
  • Resolve — see the computed value, not the expression, with a mark that says how it got there.
  • Edit — drag a wire’s source dot to re-point a reference, or type a new expression in the properties panel. It is validated as Bicep before it can touch a file.
  • Review and Save — every change shows its exact one-line diff, staged until you press Save. Only the expression you changed is rewritten; comments, alignment and ordering stay exactly as you wrote them.
  • Analyze and Show Diffs work the same way. Upgrade Preview and copy-between-canvases are Terraform-first for now.

See The canvas and Editing for the shared mechanics.

What maps to what

If you already know the Terraform experience, here is how Bicep lines up:

Terraform Bicep
.tf source files .bicep source files
.tfvars values file .bicepparam file
variable param — addressed as var.<name>
locals var — addressed as local.<name>
output output — addressed as output.<name>
a resource (aws_instance.web) a resource by its symbol (sa, or a nested child sa::blob)
simulated workspaces not applicable — use more than one .bicepparam file

Environments are .bicepparam files

A Bicep canvas does not have simulated workspaces. Instead, add more than one .bicepparam file to the canvas — each is already its own environment (dev.bicepparam, prod.bicepparam) and each is evaluated in place.

Two parameter files on one Bicep canvas. dev.bicepparam is selected and the properties panel lists the values it sets.
Two parameter files on one Bicep canvas. dev.bicepparam is selected and the properties panel lists the values it sets.

One canvas holds one language

A canvas is Terraform or Bicep, never both — the first file you add decides which. If you point one at a folder that contains both kinds, SchematIaC asks you to add them separately rather than guessing.

Modules

A local Bicep module (a referenced .bicep file) loads as a box with input and output ports, like a Terraform module. A registry module (br:) shows as a box with a note rather than real ports — there is no local vendoring step to resolve it against yet.

Local modules as boxes with ports: env from prod.bicepparam flows through the vars in main.bicep into module.storage and module.func.
Local modules as boxes with ports: env from prod.bicepparam flows through the vars in main.bicep into module.storage and module.func.

What is not evaluated yet

  • Loop resources ([for ...]) render as a single node, the same way Terraform’s count/for_each do.
  • User-defined Bicep functions (func) are not evaluated; rows that use them read unresolved rather than guess.
  • The Folder Overview is Terraform-only for now — a Bicep folder opens straight on a canvas instead.

The full list lives in Limits.

Bring your AI agent

The same Bicep engine is available to AI agents over the MCP server: your agent can read the graph, resolve a value, and get a validated Bicep diff for a proposed change — still local, still no CLI binary.