Claude Tools

A collection of command-line tools and Vim integration for interacting with Claude AI.

Overview

This repository contains several tools for using Claude AI in your daily workflow:

1. claude-ask - Command-line tool for asking Claude questions and getting responses
2. claude-bash - Generate and execute bash commands using natural language
3. claude-fix - Fix issues in scripts and files using Claude
4. claude-agent - A CLI agent that uses Claude AI to execute system tasks
5. claude-vim - Vim plugin for seamless Claude integration in your editor

All tools use the secure tulikieli credential manager to retrieve your Claude API key.

Installation

Prerequisites

• Arch Linux (or other Linux distribution with minimal adjustments)
• jq package installed (sudo pacman -S jq)
• curl package installed (usually pre-installed)
• Tulenkantajat secure credential manager set up with your Claude API key

Setting Up Your Claude API Key

This project uses Tulenkantajat, a secure GPG-based credentials management system, to safely store and retrieve your Claude API key.

# First install Tulenkantajat if you haven't already
# Follow the instructions at: https://github.com/mijuny/tulenkantajat

# Add your Claude API key to the credential manager
tulikieli add claude api_key "your_actual_api_key_here"

Installing the Tools

1. Clone this repository or copy the scripts to your preferred location
2. Make the scripts executable:

   chmod +x ~/bin/claude-ask
   chmod +x ~/bin/claude-bash
   chmod +x ~/bin/claude-fix
   chmod +x ~/bin/claude-agent.sh

3. Ensure ~/bin is in your PATH:

   echo 'export PATH="$HOME/bin:$PATH"' >> ~/.bashrc
   # Or for zsh
   echo 'export PATH="$HOME/bin:$PATH"' >> ~/.zshrc

4. Install the Vim plugin by copying it to your Vim plugin directory:

   mkdir -p ~/.vim/plugin
   cp claude-vim.vim ~/.vim/plugin/

Command-line Tools

claude-ask

Ask Claude questions and receive answers directly in your terminal.

Usage

# Basic usage
claude-ask "What is quantum computing?"

# Process content from a file
cat file.txt | claude-ask "Summarize this text"

# Save response to a file
claude-ask -o response.md "Write a markdown document about climate change"

Options

• -o FILE - Save response to the specified file
• -m MODEL - Specify Claude model (default: claude-3-7-sonnet-20250219)
• -t TOKENS - Maximum tokens in response (default: 4096)
• -v - Verbose mode (show API request details)
• -h - Show help message

Features

• Process piped input for analyzing files
• Save responses to files
• Control model parameters
• Handles special characters and formatting correctly

claude-bash

Generate and execute bash commands from natural language descriptions.

Usage

# Generate a command to list specific files
claude-bash "find all .sh files in my home directory"

# System information commands
claude-bash "show me my current disk usage"

# Network commands
claude-bash "find my ip address"

Features

• Generates bash commands based on natural language descriptions
• Shows the command and asks for confirmation before executing
• Allows editing the command before execution
• Automatically fixes common issues like unquoted wildcards
• Safety checks for potentially destructive commands

claude-fix

Fix issues in scripts and code files using Claude.

Usage

# Fix issues in a script
claude-fix script.sh "Fix the error handling in this script"

# Update a configuration file
claude-fix config.json "Add proper indentation and comments"

# Fix a script and include its runtime output for context
claude-fix script.py "The script fails when processing empty input"

Features

• Reads a file, has Claude analyze and fix it
• Creates a backup of the original file (.bak extension)
• Preserves file permissions
• Performs sanity checks before updating the file
• Prompts for confirmation for significant changes

claude-agent

A CLI agent that uses Claude AI to execute system tasks by determining and running the appropriate commands.

Usage

# Basic usage
claude-agent.sh "Collect all hardware information about my computer"

# Save output to a file
claude-agent.sh -o system_report.md "Generate a system report"

# Set a custom output directory
claude-agent.sh -d /path/to/output "Check disk usage and suggest cleanup"

Options

• -o FILE - Save final output to the specified file
• -d DIRECTORY - Output directory (default: ~/claude_agent)
• -m MODEL - Specify Claude model (default: claude-3-7-sonnet-20250219)
• -i ITERATIONS - Maximum iterations (default: 5)
• -t TIMEOUT - Command timeout in seconds (default: 30)
• -v - Verbose mode (show API requests and responses)
• -r - Force root access (use sudo for all commands)
• -h - Show help message

Features

• Intelligently breaks down complex tasks into step-by-step commands
• Creates detailed logs of all commands and their outputs
• Validates commands before execution for safety
• Handles sudo commands with user confirmation
• Timeouts for long-running commands
• Organizes outputs in session-based directories

Vim Integration (claude-vim.vim)

Seamlessly use Claude AI directly within Vim.

Commands

• :AskClaude - Ask Claude a question
• :AskClaudeSelection - Send selected text to Claude with an optional prompt
• :AskClaudeBuffer - Send the entire buffer/file to Claude for analysis

Keyboard Mappings

• <leader>ca - Ask Claude (in normal mode) or send selection to Claude (in visual mode)
• <leader>cf - Send the entire file/buffer to Claude

Usage Examples

Asking Questions

1. In normal mode, press <leader>ca
2. Type your question at the prompt
3. A new buffer opens with Claude's response

Analyzing Code

1. Select code in visual mode (v)
2. Press <leader>ca
3. Enter a prompt like "Explain this code" or "Fix these bugs"
4. Claude analyzes the selection and shows results in a new buffer

Processing an Entire File

1. With a file open in Vim, press <leader>cf
2. Enter a prompt like "Refactor this script" or "Optimize this code"
3. Claude processes the entire file and shows results in a new buffer

Moving Between Buffers

After Claude responds in a new buffer, you can navigate between buffers:

• Ctrl+o - Jump back to previous buffer
• Ctrl+i - Jump forward
• Ctrl+^ - Toggle between current and last buffer
• :bp and :bn - Previous and next buffer

Tips and Best Practices

1. For code analysis: Use specific prompts like "Find bugs" or "Optimize this function" rather than general ones.

2. For document creation: Use claude-ask -o filename.md to directly save responses to files.

3. For shell commands: When using claude-bash, review the command before execution, especially for file operations.

4. For the agent: When using claude-agent.sh, provide clear task descriptions and review any sudo commands before confirming execution.

5. Vim workflow: Use split windows (Ctrl+w s or Ctrl+w v) to view Claude's response alongside your code.

6. Security: Review all commands from claude-bash and claude-agent.sh before execution.

Troubleshooting

API Key Issues

If you see "Could not retrieve Claude API key":

tulikieli get claude api_key

If this doesn't return your API key, add it with:

tulikieli add claude api_key "your_actual_api_key_here"

If you're having issues with the Tulenkantajat credential manager:

# Check if Tulikieli is properly installed
which tulikieli

# View available credentials
tulikieli list

# Check the Tulenkantajat project documentation for more help
# https://github.com/mijuny/tulenkantajat

Command Not Found

If you see "command not found" errors, ensure the scripts are executable and in your PATH:

chmod +x ~/bin/claude-*
echo $PATH  # Verify ~/bin is included

Vim Plugin Not Working

If Vim commands aren't recognized:

:scriptnames  " Check if claude-vim.vim is listed
:source ~/.vim/plugin/claude-vim.vim  " Reload the plugin

License

These tools are provided under the MIT License.

Acknowledgments

• Developed for personal use with the Claude API by Anthropic.
• Uses Tulenkantajat for secure credential management.