- Home
- Skills
- Agents & Orchestration
- Sonariel MCP Integration
More screenshots
Works with the AI Agents you already use
Sonariel MCP Integration
- **Apply non-destructive effects:** create native DSP previews, compare results, and preserve the source recording.
Free
Sonariel MCP Integration
Example session with this skill installed
I have a 12-second WAV recording and an empty export folder on my Windows PC.
Using Sonariel, create a project called Audio Workshop and a session called First pass.
Import the WAV, tag it as
demo, and create one non-destructive
-3 dB gain preview. Tag the processed version as demo too.
Export both the original and processed versions to the folder I selected. Preserve my source file and include export metadata.
Do not install any models, generate music or speech, delete files, or automatically choose a favorite. Tell me what was completed and what I can compare.
- Read your context and instructions
- Compiled the sonariel mcp integration
Sonariel audio workflow completed
Your WAV has been imported, a non-destructive gain preview has been rendered, and both versions have been exported for comparison.
Project and results
| Item | Result |
|---|---|
| Project | Audio Workshop |
| Session | First pass |
| Original recording | Imported and preserved |
| Processed version | -3 dB native FX preview |
| Organization | Both versions tagged demo |
| Export | Original and preview WAVs exported |
| Music or voice generation | Not started |
| Additional model downloads | None |
| Favorite selection | None |
Audio processing
Original recording
- Duration: 12 seconds
- Format: Stereo WAV, 24 kHz, PCM16
- Original file: Preserved without modification
Processed preview
- Effect: Gain reduction of -3 dB
- Duration: 10 seconds
- Output: Separate derived WAV
- Source relationship: Preserved in the job/result lineage
The effects operation created a new version rather than overwriting the input. You can now listen to both results in Sonariel and decide which one to use.
Export verification
Both requested versions were exported to the authorized destination.
The export checks confirmed that the expected files were produced and that their hashes matched the recorded outputs. The portable export metadata excluded private source paths and prompt content.
No additional audio versions were generated, and no result was marked as a favorite without your choice.
What to do next
- Open the exported original and the -3 dB preview.
- Listen to the level difference and decide whether the gain reduction suits your recording.
- Keep the original as your reference. Apply further processing only if you want another version.
Validation evidence
This demonstration is based on a real isolated Codex workflow using Sonariel v0.6.1 and a synthetic WAV.
The run made
14 MCP calls across 13 tools and completed
one native FX job. An independent verification passed
27 assertions, including original-file integrity, expected audio gain, request-key consistency, result lineage, export hashes, and metadata privacy.
The measured maximum sample error against the expected -3 dB transformation was approximately 6.62 × 10^-9.
This is a formatted summary of verified test outcomes, not a verbatim agent transcript or a claim of musical quality evaluation.
Connects securely to your tools. The creator never sees your data.
About this skill
Bring local audio production into your AI agent workflow
Sonariel MCP Integration is a free skill that helps AI agents work with Sonariel, a Windows desktop application for local audio production, music generation, voice synthesis, and non-destructive effects processing.
Instead of manually navigating every technical operation, describe your audio task and let a connected agent use Sonariel's MCP tools while preserving your original files and keeping important decisions under your control.
What you can do
- Organize your audio library: create projects and sessions, import WAV files, add tags, inspect results, and manage audio versions.
- Apply non-destructive effects: create native DSP previews, compare results, and preserve the source recording.
- Export verified WAVs: export selected original and processed versions, with traceable metadata and file integrity checks.
- Work with AI music generation: discover available engines and review supported music recipes, including Stable Audio 3 Medium FP32 + SAME-L for text or source-guided generation.
- Manage voice workflows: prepare supported voice generation requests using explicitly selected languages, settings, and authorized references.
- Track persistent jobs: check real queue status and recover uncertain submissions without blindly generating duplicates.
- Manage storage safely: distinguish reversible trash, restoration, external files, and separately authorized permanent deletion.
Start without downloading an AI model
The simplest workflow only needs an existing WAV file.
Ask your agent to create a project, import your audio, apply a short gain adjustment, and export both the original and the processed preview.
Sonariel's native effects engine runs locally without installing a generative model.
Built for controlled, traceable workflows
The skill helps your agent inspect actual engine capabilities, check prerequisites, preserve original recordings, and keep track of job IDs and results.
Model installation, license acceptance, generation, and destructive storage operations remain explicit decisions. A technically valid audio result is not automatically treated as an artistically approved result.
Requirements
- Windows x64 with the free Sonariel desktop application installed separately.
- A compatible local AI agent connected to Sonariel's stdio MCP server.
- Files and output locations selected or authorized by the user.
- Additional audio models and available RAM when using generative features.
Important: This skill provides instructions, not the Sonariel application, its models, or a hosted MCP server. Audio operations execute locally, but the connected agent may use a remote AI service.
Tested compatibility
The integration was independently tested with Codex CLI using the official Sonariel v0.6.1 runtime. The v0.6.2 maintenance release preserves the same 62 MCP tool contracts.
Other agent clients have documented configuration paths but have not received equivalent end-to-end validation.
The skill is free, and the Sonariel application is available through its official GitHub repository.
- Requires Windows x64 and a separately installed Sonariel app with local MCP connectivity. Web-only agents cannot access the local server without an appropriate bridge.
- End-to-end workflow tested with Codex CLI against Sonariel v0.6.1. The v0.6.2 maintenance release preserves the MCP contracts; other agent workflows remain unvalidated.
- Generative models are optional and require separate setup, license acceptance and sufficient RAM. Medium fidelity requires 12 GiB available for admission.
- GPU inference, Stable Audio inpainting, SFT generation, packaged loop creation and active-inference cancellation are not supported in the tested distribution.
- Full Windows installer/updater lifecycle testing remains incomplete. Updater signing is separate from Windows publisher signing.
- Artistic quality, musical originality, voice consistency and third-party commercial rights are not guaranteed.
- The managed trash has a 30-day retention period, not indefinite backup. Permanent deletion requires separate authorization.
- Marketplace delivery through Agensi has not yet been validated.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 13 days ago
- Free with an Agensi account
Requires
Permissions
Sonariel runs audio processing locally through its separately installed stdio MCP server. Terminal access may be needed for local MCP setup and diagnostics; network access covers authenticated localhost communication and explicitly authorized model downloads. File operations are limited to user-selected inputs, project data, and authorized export destinations. The agent may read metadata, manage requested project operations, create previews, and export WAVs. Model installation, license acceptance, generation, and destructive operations require appropriate user authorization. Permanent deletion requires explicit selection, preview, and confirmation; external-file deletion is opt-in. Loading this skill does not authorize arbitrary shell commands, external audio uploads, access to unrelated accounts, or collection of credentials. The connected AI agent may use a remote service.