Basic Usage Guide
November 29, 2025 · View on GitHub
This guide walks you through the essential steps to get started with Claudiomiro, from installation to your first automated development task.
Table of Contents
- Installation
- Prerequisites
- Configuration
- Your First Task
- Common Workflows
- Understanding the Output
- Tips and Best Practices
- Troubleshooting
Installation
Via npm (Recommended)
npm install -g claudiomiro
Verify Installation
claudiomiro --version
Prerequisites
Required
- Node.js 18.x or higher
- Git installed and configured
- AI CLI Tool: At least one of the following:
- Claude CLI (default)
- Codex CLI
- Gemini CLI
- DeepSeek CLI
- GLM CLI
Recommended (for Best Results)
- Linting: ESLint, Pylint, or equivalent for your language
- Unit Tests: Jest, Pytest, or equivalent test framework
- TypeScript (if applicable): Configured
tsconfig.json
Having linting and tests creates a feedback loop that allows Claudiomiro to validate and fix its own work automatically.
Configuration
AI Executor Setup
Claudiomiro uses Claude as the default AI executor. Ensure you have the Claude CLI installed and authenticated:
# Install Claude CLI (if not already installed)
npm install -g @anthropic-ai/claude-cli
# Authenticate
claude login
For other AI providers, see:
Project Setup
No additional configuration is needed. Claudiomiro works with any project structure. Simply navigate to your project directory and run Claudiomiro.
Your First Task
Interactive Mode
The simplest way to start is interactive mode:
cd /path/to/your/project
claudiomiro
Claudiomiro will:
- Ask for your task description
- May ask clarification questions
- Break down the task into subtasks
- Execute tasks in parallel when possible
- Review and test the code
- Commit the changes
Direct Prompt Mode
For quick tasks, provide the prompt directly:
claudiomiro --prompt="Add input validation to the user registration form"
Example: Adding a Feature
# Navigate to your project
cd ~/projects/my-webapp
# Run Claudiomiro with a task
claudiomiro --prompt="Add a dark mode toggle to the settings page"
Claudiomiro will autonomously:
- Analyze your codebase
- Plan the implementation
- Create the necessary components
- Add styling
- Review the code
- Run tests (if available)
- Commit with a descriptive message
Common Workflows
1. Implement a New Feature
claudiomiro --prompt="Implement user authentication with JWT tokens"
2. Fix Failing Tests
When your tests are failing, let Claudiomiro fix them:
claudiomiro --fix-command="npm test"
This will:
- Run your test command
- If tests fail, analyze the errors
- Apply fixes automatically
- Repeat until all tests pass
3. Fix Linting Errors
claudiomiro --fix-command="npm run lint"
4. Security Audit
Find and fix security issues iteratively:
claudiomiro --loop-fixes --prompt="Find and fix security vulnerabilities"
5. Code Review Before PR
Get a comprehensive Staff+ level code review:
claudiomiro --fix-branch
6. Refactoring
claudiomiro --prompt="Refactor the payment processing module to use async/await"
7. Adding Tests
claudiomiro --prompt="Add unit tests for the user service module"
Understanding the Output
The .claudiomiro Folder
After running Claudiomiro, a .claudiomiro folder is created in your project root with all session files stored under task-executor/:
.claudiomiro/
└── task-executor/
├── AI_PROMPT.md # Generated AI instructions
├── CLARIFICATION_QUESTIONS.json # Questions asked (if any)
├── CLARIFICATION_ANSWERS.json # Your answers (if any)
├── CRITICAL_REVIEW_PASSED.md # Marks successful review
├── BUGS.md # Any bugs found during review
├── done.txt # Completion marker
└── TASK1/
├── TASK.md # Task description
├── TODO.md # Implementation checklist
├── RESEARCH.md # Codebase analysis
├── CONTEXT.md # Implementation guidelines
└── CODE_REVIEW.md # Review results
Log Files
For --fix-command operations:
.claudiomiro/
└── fix-command.log # Detailed execution log
Tips and Best Practices
1. Be Specific with Prompts
# Good - Specific and clear
claudiomiro --prompt="Add email validation to the registration form using regex"
# Less effective - Too vague
claudiomiro --prompt="Fix the form"
2. Start with Small Tasks
Begin with smaller, well-defined tasks to understand how Claudiomiro works before tackling complex features.
3. Review Before Pushing
Use --push=false to review changes before they go to remote:
claudiomiro --prompt="Add caching layer" --push=false
4. Use Appropriate Limits
For complex tasks, you might want more attempts:
claudiomiro --fix-command="npm test" --limit=30
For quick fixes, reduce the limit:
claudiomiro --fix-command="npm run lint" --limit=5
5. Work on the Current Branch
By default, Claudiomiro creates a new branch. To work on your current branch:
claudiomiro --same-branch --prompt="Quick fix for the login bug"
6. Run Specific Steps
If you need to re-run only certain steps:
# Re-run implementation and review only
claudiomiro --steps=6,7
7. Control Parallelism
Limit concurrent tasks if you have resource constraints:
claudiomiro --maxConcurrent=2
Troubleshooting
"Claude CLI not found"
Ensure Claude CLI is installed and in your PATH:
npm install -g @anthropic-ai/claude-cli
claude --version
"No changes detected"
This usually means:
- The task was already completed
- The prompt was too vague
- There were no files matching the task context
Try being more specific with your prompt.
"Maximum attempts reached"
The task is complex or there's a persistent issue:
- Check
.claudiomiro/folder for logs and error details - Increase the limit:
--limit=50 - Or use
--no-limit(use with caution)
"Clarification questions not answered"
If Claudiomiro asked questions and you need to continue:
- Edit
.claudiomiro/CLARIFICATION_ANSWERS.jsonwith your answers - Run:
claudiomiro --continue
Tests Still Failing After Fix
- Review the
.claudiomiro/fix-command.log - Check if the test framework is configured correctly
- Try running with more verbose output in your test command
Permission Denied Errors
Ensure you have write permissions to the project directory and Git is configured correctly:
git config user.email "your@email.com"
git config user.name "Your Name"
Next Steps
- Explore Commands Reference for all available options
- Learn about Task Executor in depth
- Set up DeepSeek or GLM as alternative AI providers
Getting Help
- GitHub Issues - Report bugs or request features
- GitHub Repository - Source code and contributions