Construct Shader Graph MCP Guidance

To safely edit Construct shader graphs via MCP, you must follow an inspection-first, minimal-mutation workflow that mandates project discovery, explicit selection, targeted inspection, atomic edits, and post-edit verification. This prevents destructive changes by ensuring every action happens through MCP tools rather than ad-hoc assumptions.

The MCP session lifecycle for Construct Shader Graph node editing requires mandatory lifecycle calls, exact MCP return values and ids, and manifest-based ambiguity resolution. This deterministic loop separates read-only inspection from side-effecting operations to reduce accidental changes during node and wire modifications.

You debug and preview Construct shader graph changes through a websocket bridge by executing manifest-driven methods to apply node edits, then immediately verifying preview outputs and errors. The workflow enforces renderer-agnostic behavior and scale-aware value guidance, such as pixelSize and texelSize, during preview debugging.

Yes, you can use MCP method execution to separate read-only inspections from side-effecting graph changes. This safety model distinguishes between calls that only observe graph state and calls that mutate nodes or wires, reducing accidental edits by enforcing explicit rules for side-effecting operations.

You resolve ambiguity in Construct Shader Graph method execution using manifests by relying on manifest-driven method execution and explicit project discovery. The guidance requires using exact MCP return values and ids, ensuring that any ambiguity in node, wire, or preview changes is resolved through the manifest before applying mutations.

The limitation of ad-hoc assumptions when editing node graphs through MCP is that they bypass the mandatory session lifecycle and safety guardrails, leading to incorrect or destructive shader-graph edits. The workflow requires explicit project selection, atomic edits, and post-edit verification to prevent accidental changes.