Convert plain English to safe shell commands using LLMs.
$ neuro "show me all python processes running on port 8000"
ℹ️ Suggested: ps aux | grep python | grep 8000
⚠️ Dry-run mode. Set NEUROSHELL_EXEC=1 to execute.
[Exec] [Edit] [Cancel]:
- LLM-powered translation — Natural language → shell commands via Groq/MiniMax
- Safety-first — Dry-run by default, confirmation prompts, blocklist enforcement
- Conversation memory —上下文 across multiple turns
- Command history — Full audit trail with timestamps
- Streaming output — Real-time command execution feedback
pip install openai groq
export GROQ_API_KEY=your_key_here
export NEUROSHELL_EXEC=1 # Enable actual execution (default: dry-run)
python3 -m neuroshell.main
neuro> find all python processes on port 8000
neuro> show disk usage sorted by size
neuro> restart all stopped docker containers
neuro> search github for repos created by amsach in the last weekneuroshell/
├── core/
│ ├── translator.py # LLM translation engine
│ ├── executor.py # Command execution + streaming
│ ├── safety.py # Blocklist + validation
│ └── history.py # Command audit log
├── main.py # CLI entry point
├── config.py # Environment + defaults
└── prompt.py # System prompts
- Blocklist —
rm -rf /,> /dev/sda, pipe secrets to netcat, etc. - Dry-run default — No command executes without explicit confirmation
- Command preview — Always shows exact command before execution
- Audit log — All commands logged with timestamp, user confirmation, output
$ neuro
╭─────────────────────────────────────────────╮
│ NeuroShell v0.1 — LLM-powered shell │
│ Mode: DRY-RUN | Model: llama-4-scout │
╰─────────────────────────────────────────────╯
neuro> find all AWS secret keys in my home directory
✅ Command: grep -r "AKIA[0-9A-Z]{16}" ~/ --include="*.env" --include="*.yaml" 2>/dev/null | head -20
[Exec] [Edit: grep -r "AKIA" ~/ | head] [Cancel]
neuro> what's using the most memory on this machine
✅ Command: ps aux --sort=-%mem | head -20
[Exec] [Edit] [Cancel]
neuro> exit
Goodbye!
MIT