Skip to main content

Kiwi AI Terminal Agent — connect your local machine to AI-powered automation

Project description

Kiwi AI CLI

Connect your local machine to AI-powered automation. The Kiwi CLI agent runs on your machine and executes terminal commands on behalf of the LLM agent via a secure WebSocket connection.

Installation

Using pipx (recommended for macOS / Linux)

pipx installs CLI tools in isolated environments — no conflicts, no venv to manage.

# Install pipx if you don't have it
brew install pipx     # macOS
sudo apt install pipx # Ubuntu/Debian

# Install kiwi-cli
pipx install kiwi-ai-cli

Using pip

pip install kiwi-ai-cli

Getting externally-managed-environment error? Modern macOS/Linux with Homebrew or system Python blocks system-wide pip installs (PEP 668). Use pipx instead, or run inside a virtual environment:

python3 -m venv ~/.kiwi-env
source ~/.kiwi-env/bin/activate
pip install kiwi-ai-cli

Windows

pip install kiwi-ai-cli

Upgrade

# If installed with pipx
pipx upgrade kiwi-ai-cli

# If installed with pip
pip install --upgrade kiwi-ai-cli

Verify your version:

pip show kiwi-ai-cli

Quick Start

# Connect to Kiwi AI (production server, restricted mode)
kiwi connect

# You'll be prompted for email and password
# Once connected, the LLM agent can run commands on your machine

Usage

# Connect to production (default)
kiwi connect

# Connect to a specific server
kiwi connect --server dev
kiwi connect --server local
kiwi connect --server wss://custom.server.com

# Provide email upfront (password is always prompted securely)
kiwi connect --email you@example.com

# Skip login with an existing token
kiwi connect --token <your-jwt-token>

# Allow access to additional directories
kiwi connect --allow ~/Documents --allow ~/Projects

# Full access mode (no path restrictions)
kiwi connect --scope full

Flags

Flag Description
--server Server preset (app, dev, local) or a full WebSocket/HTTP URL. Default: app (production)
--email Email address for login. Prompts interactively if not provided
--token Access token to skip email/password login (e.g., from shared auth)
--scope restricted (default) limits commands to the current directory. full removes all path restrictions
--allow Additional allowed directory path. Can be used multiple times. Only applies in restricted mode

Execution Modes

Restricted Mode (default)

Commands are validated before execution. The LLM can only access files within:

  • The directory where you ran kiwi connect
  • Any directories added via --allow

Paths like /dev/null, /dev/stdin, etc. are always allowed.

Full Mode

No path validation. The LLM can execute any command. Use with caution.

kiwi connect --scope full

Server Presets

Preset URL
app (default) wss://api.meetkiwi.ai
dev wss://dev.api.myautobots.com
local ws://localhost:8000

You can also pass a full URL:

kiwi connect --server https://my-server.com

How It Works

  1. You run kiwi connect on your machine
  2. The CLI authenticates and opens a WebSocket connection to the server
  3. When the LLM needs to run a command, the server sends it through the WebSocket
  4. The CLI executes it locally and streams the output back
  5. The LLM sees the output and decides what to do next

Commands can run in one-off mode (independent, no state) or in persistent sessions (shell state like cd, export, environment variables carry over between commands).

Requirements

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

kiwi_ai_cli-0.2.0.tar.gz (14.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

kiwi_ai_cli-0.2.0-py3-none-any.whl (15.1 kB view details)

Uploaded Python 3

File details

Details for the file kiwi_ai_cli-0.2.0.tar.gz.

File metadata

  • Download URL: kiwi_ai_cli-0.2.0.tar.gz
  • Upload date:
  • Size: 14.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.11

File hashes

Hashes for kiwi_ai_cli-0.2.0.tar.gz
Algorithm Hash digest
SHA256 67ac23f3de914da5e53ad68c37a3971f50bb2c83102c847813ecbc2db88ecd98
MD5 e6605636ffada477da678fb8051f76bf
BLAKE2b-256 62add913ff53b0f42255538dc0ad93d9a35c62958e5176d531aef3af8bfb2675

See more details on using hashes here.

File details

Details for the file kiwi_ai_cli-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: kiwi_ai_cli-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 15.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.11

File hashes

Hashes for kiwi_ai_cli-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 242b170bed170e7c1838dea1cc384eef6db66188e8563139261e63b2f3cd4deb
MD5 f897c9c2ff4823d4bb3b46336f0a72b7
BLAKE2b-256 eb6c2928114fa736d5279a405f23a08acaa74544bb875f9506449a8d8619e313

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page