Skip to main content

AI-powered command line assistant - any model, anywhere

Project description

MaShell

███╗   ███╗ █████╗ ███████╗██╗  ██╗███████╗██╗     ██╗     
████╗ ████║██╔══██╗██╔════╝██║  ██║██╔════╝██║     ██║     
██╔████╔██║███████║███████╗███████║█████╗  ██║     ██║     
██║╚██╔╝██║██╔══██║╚════██║██╔══██║██╔══╝  ██║     ██║     
██║ ╚═╝ ██║██║  ██║███████║██║  ██║███████╗███████╗███████╗
╚═╝     ╚═╝╚═╝  ╚═╝╚══════╝╚═╝  ╚═╝╚══════╝╚══════╝╚══════╝

🐚 Your AI-Powered Command Line Assistant

MaShell is an AI agent that lives in your terminal. It can run commands, write code, create projects, and accomplish complex tasks — all through natural conversation.


⚡ Quick Start

# Install (macOS/Linux)
pipx install mashell

# Or with pip (in a virtual environment)
pip install mashell

# Run (first time will guide you through setup)
mashell

Note: On macOS, we recommend pipx for installing CLI tools. Install it with brew install pipx if you don't have it.

That's it! MaShell will walk you through connecting to your AI provider (OpenAI, Azure, Anthropic, or local Ollama).


� What Can MaShell Do?

Task Example
Find files "find the largest video on my computer"
Read documents "read my resume.pdf and summarize it"
Edit Word files "update my resume.docx with new job title"
Create projects "build a Python web scraper from scratch"
Run code "run this script and fix any errors"
Git operations "commit all changes with a good message"
Web scraping "scrape the HN front page for top stories"
Install tools "install ffmpeg and convert video.mp4 to gif"
System tasks "clean up files older than 30 days in Downloads"
Debug "why is my Python script failing?"

Native File Tools (No Shell Required)

MaShell has built-in tools for file operations — faster and no permission prompts for reading:

Tool Description Permission
read_file Read any file (supports PDF, Word .docx, text) Auto ✅
list_dir List directory contents with sizes Auto ✅
search_files Search/grep across files Auto ✅
write_file Write text files Requires approval
edit_docx Edit Word documents (find/replace, update paragraphs) Requires approval
shell Execute any shell command Requires approval
crawl Scrape web pages (JS rendering) Requires approval

🔧 Setup Options

Option 1: Interactive Setup (Recommended)

Just run mashell — if no config exists, it will guide you:

🐚 MaShell Configuration Wizard

Step 1: Choose your LLM provider
  1. openai    - OpenAI API (GPT-4o, etc.)
  2. azure     - Azure OpenAI Service
  3. anthropic - Anthropic API (Claude)
  4. ollama    - Local Ollama (no API key needed)

Select provider: 2

Step 2: API Endpoint URL
Enter your Azure OpenAI endpoint: https://your-resource.openai.azure.com/

Step 3: API Key
Enter your API key: ********

Step 4: Model / Deployment Name
Deployment name: gpt-4o

✅ Configuration saved!

Option 2: Command Line Arguments

# OpenAI
mashell --provider openai --url https://api.openai.com/v1 --key sk-... --model gpt-4o "your task"

# Azure OpenAI
mashell --provider azure --url https://your-resource.openai.azure.com --key xxx --model gpt-4o "your task"

# Anthropic
mashell --provider anthropic --url https://api.anthropic.com --key sk-ant-... --model claude-sonnet-4-20250514 "your task"

# Local Ollama (no key needed)
mashell --provider ollama --url http://localhost:11434 --model llama3 "your task"

Option 3: Environment Variables

export MASHELL_PROVIDER="openai"
export MASHELL_URL="https://api.openai.com/v1"
export MASHELL_KEY="sk-..."
export MASHELL_MODEL="gpt-4o"

mashell "your task"

� See It In Action

Example 1: Find the Largest Video File

You: find biggest video file on my computer

💭 I'll check common video locations first...

▶ Run:
  $ du -sh ~/Documents ~/Downloads ~/Desktop ~/Movies

📋 Output:
342G    /Users/jacob/Documents
5.4G    /Users/jacob/Downloads
4.8G    /Users/jacob/Desktop
12K     /Users/jacob/Movies

💭 Documents is huge. Let me search for videos there...

▶ Run:
  $ find ~/Documents -type f \( -iname "*.mp4" -o -iname "*.mkv" \) -exec ls -lhS {} + | head -10

📋 Output:
-rw-r--r--  3.8G  Peaky Blinders S06E06.mkv
-rw-r--r--  3.7G  The.Brutalist.2024.mp4
-rw-r--r--  3.0G  Mission.Impossible.mkv
...

MaShell: Found it! The biggest video is:
**3.8G** — ~/Documents/Bitcomet/Peaky Blinders S06E06.mkv

Example 2: Build a Project From Scratch

You: create a python project that can scrape web pages

💭 I'll set up a proper Python project structure...

▶ Run:
  $ mkdir -p python-scraper/{src,tests,outputs}
✓ Done

▶ Run:
  $ cat << 'EOF' > python-scraper/src/fetch.py
  import httpx
  def fetch(url): ...
  EOF
✓ Done

... (creates requirements.txt, README, .gitignore, venv, installs deps)

▶ Run:
  $ ./run.sh scrape https://example.com
📋 Output:
URL: https://example.com
Status: 200
Title: Example Domain

MaShell: Done! Created a Python web scraper at ./python-scraper

Try it: ./run.sh scrape https://example.com --out outputs/page.html

Example 3: Scrape a Website

You: use it to scrape www.cnn.com and see what's happening today

▶ Run:
  $ ./run.sh scrape https://www.cnn.com --out outputs/cnn.html
📋 Output:
Status: 200
Title: Breaking News, Latest News and Videos | CNN
Saved: outputs/cnn.html

💭 Let me extract the headlines...

▶ Run:
  $ python -c "from bs4 import BeautifulSoup; ..."

MaShell: Today's top story: "Trump ramps up rhetoric on ICE despite calls for de-escalation"

🔐 Permission System

MaShell always asks before running commands:

╭─────────────────── 🔐 Permission Request ───────────────────╮
│                                                              │
│  Tool: shell                                                 │
│                                                              │
│  Command:                                                    │
│  $ rm -rf ./node_modules                                     │
│                                                              │
╰──────────────────────────────────────────────────────────────╯
   Approve   Deny   Edit   Always approve this session
Choice [y/n/e/a]: 
  • y — Run this command
  • n — Block it
  • e — Edit the command first
  • A — Auto-approve all commands this session (use with caution!)

💡 Tips

  1. Be specific — "find large files" → "find video files larger than 1GB in Downloads"

  2. Let it explore — MaShell thinks step-by-step like a human, checking results before deciding what to do next

  3. Auto-approve mode — Use mashell -y "task" for trusted, quick tasks

  4. Interactive mode — Just run mashell to chat back and forth


📦 Installation from Source

git clone https://github.com/anthropics/MaShell.git
cd MaShell
pip install -e .

For development:

python -m mashell "your task"

📄 License

GPL-3.0 — Free to use, modify, and share.


Built for developers who live in the terminal 🐚

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

mashell-0.1.5.tar.gz (69.0 kB view details)

Uploaded Source

Built Distribution

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

mashell-0.1.5-py3-none-any.whl (72.5 kB view details)

Uploaded Python 3

File details

Details for the file mashell-0.1.5.tar.gz.

File metadata

  • Download URL: mashell-0.1.5.tar.gz
  • Upload date:
  • Size: 69.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for mashell-0.1.5.tar.gz
Algorithm Hash digest
SHA256 47764264f4363d006e51a09e57b7233e6c565c941aae0c59ca480f82a9fdd0ea
MD5 c54a4b2d3f759a4a4640e028d74143bf
BLAKE2b-256 7fe2f9b788083f3183daf9920d88cca83b4d0799fed5bfbfcbbbe404276f1b30

See more details on using hashes here.

Provenance

The following attestation bundles were made for mashell-0.1.5.tar.gz:

Publisher: publish.yml on jacobjiangwei/MaShell

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file mashell-0.1.5-py3-none-any.whl.

File metadata

  • Download URL: mashell-0.1.5-py3-none-any.whl
  • Upload date:
  • Size: 72.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for mashell-0.1.5-py3-none-any.whl
Algorithm Hash digest
SHA256 283c69f49897c0428118c40ed4350bd686163dcd5bf345d98e240ddfe55eaeb7
MD5 83c8c696c02fdeb9eb9e9014115a5ebc
BLAKE2b-256 b56cd1f019216102f7a57c3e6e1700cee4000f95d8c2042c0823824eb69321fc

See more details on using hashes here.

Provenance

The following attestation bundles were made for mashell-0.1.5-py3-none-any.whl:

Publisher: publish.yml on jacobjiangwei/MaShell

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

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