Loomle AI builds SAL, Loomle MCP, and OASIUM—Structured Agent Language, an Unreal Engine interface, and an immersive world where humans and AI agents create together.
eventGraph = target {
domain: graph,
asset: "/Game/Blueprints/BP_Door.BP_Door",
blueprintId: "7c3f91a4-2d68-4f0e-b517-9a42e6c8d35b",
id: "e2a56b9d-73c1-48f4-8d20-c6b91a374ef2"
}
patch eventGraph dry run
print = { palette: "returned-palette-entry-id" }
add print @source-node-guid/source-pin-guid -> print.execute
move print to (640, 320)Structured Agent Language (SAL) expresses object structure, relationships, capabilities, and edits as compact, ordered, copyable text while preserving native names, values, identities, and semantics. Humans and agents can inspect, exchange, and modify the same text; computers can validate and execute it. Its compact form reduces total token cost across the full agent loop—from discovery and reading through reasoning, modification, and verification.
beginPlay = node {
id: "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa",
type: "/Script/BlueprintGraph.K2Node_Event"
}
beginPlay.then -> validate.execute / then -> unlock.execute / then -> open.execute / then -> notify.execute
# excerpt: Pin and intermediate Node bindings omittedquery eventGraph
exec flow from pin @node-guid/pin-guid depth 8patch eventGraph dry run
print = { palette: "returned-palette-entry-id" }
add print @source-node-guid/source-pin-guid -> print.execute
move print to (640, 320) SAL gives agents one compact object interface to Unreal Engine instead of adding another MCP tool for every capability. sal_schema, sal_query, and sal_patch carry discovery, reading, and editing across every supported Unreal object.
Resident Skills are discovered and loaded through MCP. They teach agents how to combine the three core calls with domain-specific judgment—from reading exact pin geometry to dry-running, moving, and verifying a Blueprint graph. The comparison below uses the same nodes and links; only their layout changes.
