Open source · MIT License · v2.0

Voice-first AI for your terminal

A standalone CLI assistant with voice input/output, persistent memory, tool calling, and live voice conversation — all from your terminal.

Python 3.9+ · macOS, Linux, Windows · One-command setup

Scroll
Install

One command. Everything set up.

Auto-installs Python deps, Ollama, DeepAgent sidecar, pulls default model, and starts all services.

$ curl -fsSL https://raw.githubusercontent.com/astromanish/aishe-cli/main/setup.sh | bash
Features

What it does

Six core capabilities. No filler.

Chat

One-shot, streaming, or REPL. Tools fire automatically.

Live voice

Press Enter. Speak. Hear the answer back.

Memory

Remembers facts across sessions. Searches before answering.

Threads

Parallel conversations. Full history. Exportable.

Voice I/O

Transcribe any file. Synthesize any text. 10 voices.

Doctor

One command checks everything. Roundtrip tests included.

Tools

Built-in tools

Five tools, loaded by default. Add your own in tools.py.

Live voice

VAD detects speech boundaries. Records, transcribes, thinks, speaks back — automatically.

Architecture

Four local services

Everything runs on your machine. No cloud, no API keys, no data leaving your computer.

DeepAgent

LangGraph + tools

Ollama

LLM inference

Parakeet STT

Speech → text

Supertonic TTS

Text → speech

CLI

Every command

15 subcommands. Shell completions for bash, zsh, and fish.

$ aishe status
$ aishe chat "What is 15 * 4?"
$ aishe stream "Tell me a joke"
$ aishe repl
$ aishe live
$ aishe memory add "User prefers Hindi"
$ aishe memory search "Hindi"
$ aishe threads --new
$ aishe voice transcribe recording.wav
$ aishe voice speak "Namaste"
$ aishe ollama models
$ aishe doctor
$ aishe search "pizza"
$ aishe export
$ aishe config set voice.default_voice M1
$ aishe completions zsh

Built in India, for India

Open source under MIT. Clone it, run it, break it, improve it.

Get started

© 2026 Aishe. MIT License. Made in India.