voxagent
Voice terminal agent that records speech, transcribes locally, and answers via Ollama.
Tested on macOS and Linux
Use it
macOS arm64
You need first
- Runtimes:
Node.js 18.3 or later, except Node.js 19 - System packages:
libvulkan1 libgomp1 libasound2t64
install voxagent via npm install -g voxagent
node --version && npm --version && npm install -g voxagent
Check that it works
voxagent --helpThe output should contain “Usage: voxagent [options]”.
First thing to try
voxagent --list-devicesRun voxagent --list-devices to see which audio input devices are available.
Linux x86_64
You need first
- Runtimes:
Node.js 18.3 or later, except Node.js 19 - System packages:
libvulkan1 libgomp1 libasound2t64
Install voxagent via npm global
npm install -g voxagent
Check that it works
voxagent --helpThe output should contain “Voice-powered terminal agent.”.
First thing to try
voxagent --file question.wav --device defaultRun voxagent --list-devices to see which audio input devices are available.
What it does
- Ask questions by speaking in the terminalNot tested
- Turn speech into text on your own machineNot shown working
- Get answers from a local modelNot tested
- Ask follow-up questions in the same conversationNot tested
How far pickfit got
macOS arm64
- Installs
- Starts
- 0 of 4 features work
- Works offline
Linux x86_64
- Installs
- Starts
- Main function works
- Works offline
Other systems
Windows: the project says it runs there; pickfit hasn’t tried it.
Good for
- ask a spoken question from the terminal
- answer one question recorded in an audio file
- ask follow-up questions with conversation memory
- list available audio input devices
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
macOS arm64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 1 Nov 2026
- Version 0.3.0
Differences the test machine hit
- README: README lists Linux prerequisites libvulkan1 libgomp1 libasound2t64 and Windows VC++ redistributable
In the test: On macOS 14+ Apple Silicon no extra system libraries needed; npm install succeeded with Node 26.13.0 - README: voxagent --file described as loading Ollama and whisper model before transcribing
In the test: voxagent reads file and checks for speech first, before contacting Ollama or loading whisper; verified with silence file returning 'No speech detected' without Ollama running
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 0.3.0
Differences the test machine hit
- README: On Debian and Ubuntu, `sudo apt install libvulkan1 libgomp1 libasound2t64`
In the test: In this box (no root, no apt) libasound.so.2 is missing, so `voxagent --file /tmp/does-not-exist.wav` exits 1 with 'A native module could not load because a library it needs is missing ... install it with: sudo apt install libasound2t64' instead of a file-read error; --help and --file/--device vali…