AI Agents & DSL Console

Chat agents, a DSL REPL, and per-device sound design. Cloud or local inference.

MAGDA AI chat panel — The AI chat panel — natural-language requests turn into auditable DSL commands.

MAGDA ships with AI agents that have access to your session context and act on it by emitting commands in a built-in domain-specific language. You can use the chat, or type commands directly in the REPL.

New in 0.7: per-device sound-design agents. The 4OSC has its own AI panel docked next to the synth — type a description of the sound, and the model streams back a preset and applies it directly. The preset name and category come along too, so saving into the new preset system is two clicks. More devices to follow.

Cloud providers such as OpenAI, Anthropic and Google require your own account and API credentials. Local inference via llama.cpp or a local server is also supported.

For what to expect from the AI and how to get the most out of it, see AI in MAGDA: Philosophy and Expectations in the manual.

Agents operate on real project state, not a preview
DSL REPL for direct, auditable commands
Per-device sound-design agents starting with 4OSC (new in 0.7)
Switch between cloud and local models from settings

MAGDA DSL console — The DSL console — type commands directly for precise, scriptable control.