Schema Reference

rune.schema.json is the canonical binding manifest — the artifact that turns cross-layer contracts from aspirational to enforceable.

Without the schema, Rune is a convention. With it, Rune is a contract.

The schema is hosted at rune.semanticintent.dev/rune.schema.json (JSON Schema draft-07).

---

Manifest structure

{
  "$schema": "https://rune.semanticintent.dev/rune.schema.json",
  "$rune": "1.1",
  "host": {
    "format": "react",
    "version": "1.0.0",
    "source": "app.tsx"
  },
  "bindings": {
    "risk-threshold": {
      "rune": "~",
      "type": "number",
      "min": 0.05,
      "max": 0.30,
      "intent": "approved by risk committee Q1-2025 — review at quarter end"
    },
    "market-price": {
      "rune": "@",
      "type": "number",
      "intent": "live NBBO feed — not manually entered"
    },
    "submit-order": {
      "rune": "!",
      "args": [{ "name": "order-id", "type": "string", "required": true }],
      "intent": "explicit, logged to OMS, irrevocable"
    },
    "screen-purpose": {
      "rune": "?",
      "intent": "trading terminal — real-money consequences, every action logged"
    }
  }
}

---

Top-level fields

Field	Required	Description
$schema	No	URI to the meta-schema — enables IDE validation
$rune	Yes	Protocol version — must be "1.1"
host	No	Metadata about the host that owns these bindings
bindings	Yes	Named binding declarations (see below)

host

Field	Description
format	Host format identifier — "mere", "react", "csharp", "sql", etc.
version	Version of the host document or application
source	Path or identifier of the source document

---

Binding declaration

Binding keys must be kebab-case (^[a-z][a-z0-9-]*$). This is the canonical form — hosts translate to their own conventions.

Host	risk-threshold becomes
SQL	risk_threshold
C#	RiskThreshold
React	'risk-threshold'

The schema is the arbiter when layers disagree on naming.

Binding fields

Field	Required	Description
rune	Yes	"@" read, "~" sync, "!" act, "?" intent
type	Recommended	Value type — string, number, boolean, array, object
intent	Required for ?	The ? annotation — rationale, constraint, governance
args	For ! only	Parameter declarations (see below)
min	For number	Minimum value — RNE005 if violated
max	For number	Maximum value — RNE005 if violated
enum	For string	Allowed values — RNE005 if an unlisted value is assigned
maxLength	For string	Max character length — maps to PIC X(n) in RECALL, VARCHAR(n) in SQL

Rune types

Rune	Name	What it governs
@	read	One-way, state → display. No side effects. Not writable.
~	sync	Bidirectional. Input and state stay in sync. Mutable.
!	act	Explicit invocation. Nothing happens without a call.
?	intent	Annotation only. No runtime effect. Travels with the binding.

args (for ! act bindings)

"submit-order": {
  "rune": "!",
  "args": [
    { "name": "order-id",  "type": "string",  "required": true  },
    { "name": "quantity",  "type": "number",  "required": true  },
    { "name": "dry-run",   "type": "boolean", "required": false }
  ]
}

Field	Description
name	Parameter name
type	string, number, boolean, state-ref
required	Whether the parameter must be supplied at call sites

state-ref means the argument value is a binding id — the action receives the current state value of that binding.

---

Constraint validation (RNE005)

The schema validator raises RNE005 when:

• min or max is declared on a non-number binding
• enum is declared on a non-string binding
• min is greater than max
• args is declared on a non-! binding

---

Complete example

A Mere workbook binding manifest showing all four runes with full constraint declarations:

{
  "$schema": "https://rune.semanticintent.dev/rune.schema.json",
  "$rune": "1.1",
  "host": { "format": "mere", "source": "task-workbook.mp.html" },
  "bindings": {
    "tasks":           { "rune": "@", "type": "array",  "intent": "computed — do not mutate directly" },
    "new-task":        { "rune": "~", "type": "string", "maxLength": 200, "intent": "cleared after !add-task fires" },
    "task-filter":     { "rune": "~", "type": "string", "enum": ["all", "active", "completed"], "intent": "controls @tasks subset" },
    "task-count":      { "rune": "@", "type": "number", "intent": "live count of active tasks — display only" },
    "add-task":        { "rune": "!", "args": [{ "name": "title", "type": "state-ref", "required": true }], "intent": "no-op if ~new-task is empty" },
    "complete-task":   { "rune": "!", "args": [{ "name": "task-id", "type": "string", "required": true }], "intent": "irreversible" },
    "clear-completed": { "rune": "!", "intent": "destructive — no undo" },
    "workbook-purpose":{ "rune": "?", "intent": "personal task tracking — single user, no persistence beyond session" }
  }
}

See examples/task-workbook.rune.json in the repository.

---

Tooling

Use rune validate from @semanticintent/rune-cli to validate any manifest against this schema:

rune validate my-app.rune.json
rune validate my-app.rune.json --format json   # CI-friendly

Use rune extract to generate a manifest from existing source:

rune extract src/          # scan and bootstrap
rune validate rune.json    # see what's missing

See CLI reference →