Discover first
Agents can inspect available models and voices before submitting work to capabilities that are not ready.
Sonano × Your agent
Give your agent
a local voice.
Let Sonano on your Mac handle speech, transcription and voice conversion. Your agent organizes the work; you can review progress and results in the same desktop workspace.
Workflow illustration · No live execution
Your agentUnderstand · Orchestrate
SonanoLocal audio · Traceable results
Connect. Then create.
Stay in your own workspace.
The app provides setup instructions for your host and version. Once connected, your agent and the GUI share models, voices, jobs and artifacts.
Use Sonano in Cursor
Choose Cursor in the app’s Connect Agent page to bring local voice tools into your workspace.
This is a setup checklist. Copy actual commands and configuration from the app to avoid outdated or unauthenticated examples.
- Open Connect Agent in Sonano and choose Cursor.
- Follow the app’s instructions to install the Skill or configure MCP, then enable the server in Cursor when required.
- Ask your agent to check available capabilities, test a short passage and inspect the result.
Use Sonano in Claude Code
Hand audio work to the local engine and follow progress and results from your agent environment.
This is a setup checklist. Copy actual commands and configuration from the app to avoid outdated or unauthenticated examples.
- Open Connect Agent in Sonano and choose the appropriate host setup.
- Configure MCP or install the relevant Skill using the instructions currently shown in the app; restart the host if needed.
- Confirm that tools and local models are available before requesting speech, transcription or voice conversion.
Use Sonano in Codex
Reusable voices, jobs and artifacts follow the same local workflow.
This is a setup checklist. Copy actual commands and configuration from the app to avoid outdated or unauthenticated examples.
- Choose Codex or the shared Skill option in Sonano’s Connect Agent page.
- Use the Skill installation or MCP setup instructions provided by the app, then enable them in your host.
- Ask the agent to discover capabilities, choose an available voice and verify a short task.
Use the agent you know
OpenClaw, Hermes, scripts and compatible hosts can use the same local voice capabilities.
This is a setup checklist. Copy actual commands and configuration from the app to avoid outdated or unauthenticated examples.
- Choose your host in Connect Agent, or use the shared Skill / MCP instructions.
- Complete the setup required by your host. For HTTP calls, use the authenticated examples provided by the app.
- Run the connection test before media work. Access to external files depends on host permissions.
From request to result
Hand off the task.
Keep the result.
Long-running work has progress; results have a source. After the agent gets the output, you can review history, reuse voices and export from Sonano.
Fixed example · Explore each step
One request, several steps.
Transcribe the riverside interview and export English subtitles. Then use a warm, measured narrator to read a short introduction.
The agent discovers available capabilities, models and voices before submitting work. This example explains the flow; it does not execute in your browser.
Every step has a visible state.
Interview transcriptionExample: completed
Spoken introductionExample: completed
In actual use, Jobs & Progress shows queued, processing, completed, canceled or failed work. The agent and app read the same job records.
Results you can keep working with.
SRT
riverside-interview.srtExample subtitle artifact · Review and export
WAV
spoken-introduction.wavExample audio artifact · Return to the workspace
Real jobs return queryable artifact references. These filenames are illustrations, with no fabricated downloads.
One local system
Automation you can follow.
Connection is part of the workflow. Model readiness, voice sources and permissions matter just as much when an agent is doing the work.
Trace the work
Speech, transcription and conversion use the shared job system. Inspect status, errors and the resulting artifacts.
Keep the same boundaries
The service listens on 127.0.0.1 by default. Except health checks, all APIs require same-user local authentication. Use the app’s setup flow without recurring manual pairing.
One workspace. Two ways to create.
Edit by hand or ask your agent. Work with the same voices and results.