Open Source MCP-Wwise Server Demo: Create Project From Prompt

I open-sourced my MCP server for Wwise... 💻

https://github.com/john-tennant-sjs/mcp_wwise

TL;DR This lets your Ai agent speak directly to Wwise via WAAPI.

It has 46 validated tools that any MCP-compatible AI host (Claude, ChatGPT, Cursor... etc) can call to create, inspect, and modify Wwise project content. Every WAAPI command has its own MCP tool. There are also some additional helper tools (eg. guid/name lookup and set RTPC).

Why did I make this? WAAPI is powerful but unforgiving—silent failures, opaque error codes, tribal knowledge payload quirks... etc. So the core idea of this MCP server is a 'content-authoring-first' design through JSON schema validation. Every tool has an input contract, an output contract, and semantic preflight checks. Malformed requests fail fast with actionable errors before they ever reach Wwise. Display names get resolved to proper WAAPI identifiers (types/objects/properties... etc.) automatically, so agents don't burn tokens trying to figure out that when a user says "RTPC", in WAAPI-script-world they really mean "GameParameter". It's like a built-in spell-checker for WAAPI commands.

The linked demo video is Sonnet 4.6 building an entire 1st Person Shooter Wwise project from scratch without any WAAPI errors. This matters because WAAPI errors can leave the Wwise application in a bad state, causing sluggish behavior and eventually an application restart. This MPC server minimizes WAAPI errors.

The server is MIT-licensed, runs locally over stdio (no network exposure), and has been extensively tested with Wwise 2023.1 and 2025.1.

If you're new to MCP servers, setup is easy for your Ai to figure out if you just tell it you want to install this and give it the link above.

#gameaudio #wwise #mcp #opensource #gamedevelopment

Видео Open Source MCP-Wwise Server Demo: Create Project From Prompt канала John Tennant