Roles & Access
SK Wwise MCP supports role-based access control by enabling different combinations of servers per user. This lets teams give AI capabilities that match each person’s responsibilities — junior sound designers get read-only access, while build engineers get full pipeline control.
Recommended Roles
| Role | Servers | What They Can Do |
|---|---|---|
| Sound Designer (Junior) | Browse, Audition, Media Read | Explore the project, preview sounds, inspect audio sources |
| Sound Designer (Senior) | + Objects, Containers | Create and edit objects, configure containers and routing |
| Build Engineer | + Pipeline, Command Line | Import audio, generate SoundBanks, run CLI operations |
| QA / Profiling | + Profiling, Profiling Control, Remote | Profile performance, connect to devkits, capture sessions |
| Admin | All servers | Full access to every tool |
How to Configure
When running setup.bat or setup.sh, you’ll be prompted to select which servers to enable. The script generates a .mcp.json file with only the selected servers.
To change the configuration later, edit .mcp.json directly — add or remove server entries as needed.
Why Split Into Servers?
Two reasons:
Safety — Fewer tools means fewer ways for an AI agent to make unintended changes. A junior designer’s agent literally cannot delete objects if the Objects server isn’t enabled.
Accuracy — LLMs select tools more reliably from smaller sets. 12 focused servers with ~8 tools each outperform a single server with 95 tools.