hashcli 0.5.0

Hash - Intelligent CLI system with dual-mode functionality (LLM chat and command proxy)

Hash CLI

Hashcli is an intelligent terminal assistant that combines LLM-powered conversational assistance with command-line productivity tools. Ask questions in plain English, manage your development workflow, and keep your shell flow fast without losing control or safety.

Why Hash CLI

• Conversational AI: Natural language interface for coding help, debugging, and system administration.
• Provider choice: OpenAI, Anthropic, and Google models supported.
• Tool calling with guardrails: shell execution, file ops, web search, and code analysis with confirmations and policy checks.
• Conversation history: list, inspect, and clear past sessions.
• Shell integration: optional # prefix for a seamless inline workflow.
• Rich output: streaming, panels, and quiet mode when needed.

Quickstart

Install Hashcli:

pipx install hashcli

or

uv tool install hashcli

You can also use pip install hashcli, but we recommend using pipx or uv tool as it makes hashcli available on all virtual environments.

Set up Shell integration:

hashcli --setup

Set up one of the following api keys in your terminal:

export GOOGLE_API_KEY="your-key"
export OPENAI_API_KEY="your-key"
export ANTHROPIC_API_KEY="your-key"

If you prefer a guided setup:

hashcli --config

Usage

# show me the current disk usage in human readable format
# show last 5 git commit messages
# how to find all __pycache__ in current folder

Features (from the current code)

• Core slash commands: /help, /history
• Plugin slash commands: install with hashcli --add-cmd <path-to-plugin.py> (installed under ~/.hashcli/plugins)
• LLM Tools:
  • execute_shell_command (guarded shell execution with timeouts and allow/block lists)
  • read_file, write_file, list_directory (filesystem operations with safety checks)
  • web_search (DuckDuckGo via ddgs)
  • analyze_code (AST-based Python analysis and lightweight JS/Java metrics)
• History management: list sessions, show a session, clear old or all history
• Streaming output: optional streaming for responses
• Interactive config wizard: hashcli --config for provider, model, and API key setup

Plugin example:

hashcli --add-cmd plugins/model.py
hashcli /model list

Configuration

Hash CLI loads configuration in this order (highest to lowest):

1. CLI flags
2. Environment variables (HASHCLI_ prefix)
3. User config (~/.hashcli/config.toml)
4. System config (/etc/hashcli/config.toml on Unix)
5. Defaults

Provider key fallbacks:

• OPENAI_API_KEY
• ANTHROPIC_API_KEY
• GOOGLE_API_KEY or GEMINI_API_KEY

Useful flags:

hashcli --model gpt-5-mini
hashcli --provider anthropic
hashcli --no-confirm
hashcli --quiet
hashcli --show-config

Safety Notes

• Shell execution and file writes require confirmation by default.
• LLM tool calls are checked against allow/deny lists and dangerous patterns.
• Shell operators (| and ;) are blocked unless explicitly enabled.