Audio editing with an automation layer
Audionaut has introduced an open-source desktop application for multitrack audio editing and recording that can also accept instructions from AI agents. The project runs on Windows, macOS and Linux, and its repository describes a workflow that joins a graphical editor with automation through the Model Context Protocol, or MCP.
The software is aimed at music, podcast and multitrack projects. Its listed editing tools include precise cuts, per-track playlists, multichannel support and exports. Audionaut is written in modern C++ using the JUCE application framework, positioning it as a native desktop program rather than a browser-only service.
MCP support is the project's distinguishing feature. With Audionaut open, compatible agents can send edits to the active session, and each received change is recorded as a separate undo step. That arrangement is designed to make automated changes inspectable and reversible inside the same project used for manual work. The repository specifically describes compatibility with Claude, while MCP is intended as a broader interface rather than a single-vendor integration.
One project format for interface and scripts
Audionaut also ships with a command-line tool for headless work. The CLI can create projects, import material, run available analysis, assemble edits and export results without opening an audio device or graphical interface. Projects written by the CLI can be opened in the desktop application and vice versa, allowing scripts and people to work on the same files.
Commands offer JSON output with a consistent success-or-error envelope, while logging is sent separately. That detail is useful for continuous integration and agent workflows because another program can parse results without trying to distinguish status messages from structured output. The project says analytics are opt-in and provides an environment setting to disable CLI reporting regardless of the preference stored by the desktop app.
Some advanced functions depend on optional components. Audio analysis such as beat tracking, onset detection and segmentation uses Essentia when it is built into the application. Stem separation uses a C++ port of Meta's Demucs and requires separately downloaded model weights. Builds without the analysis dependency can still run, but those features are unavailable.
The repository also documents platform-specific constraints. On macOS, the sandboxed application can access only permitted locations for headless operations, while the standalone CLI has fewer restrictions. Windows users are advised to prefer the CLI for shell automation because the graphical application may return control before output is complete.
Audionaut is offered under GPL version 3 or later alongside a commercial licensing option. Its launch is best understood as an early open-source product release, not an independently tested claim of professional performance. The practical proposition is clear: familiar multitrack editing remains available through a desktop interface, while structured commands give scripts and agents a controlled path into the same session.
