Terminal assistant that combines AI chat with command execution. Supports Ollama and LM Studio.
pip install xandai-cli
xandai --auto-detect# Terminal commands
xandai> ls -la
xandai> git status
# AI questions
xandai> How do I optimize this code?
# Natural conversation for any task
xandai> create a REST API with authentication- Ollama - Local models
- LM Studio - GUI-based model management
xandai --provider ollama
xandai --provider lm_studio --endpoint http://localhost:1234/agent <instruction> # Multi-step LLM orchestrator for complex tasks
/set-agent-limit <n> # Set max LLM calls (default: 20, max: 100)
/review # AI-powered code review
/web on # Enable web content integration
/help # Show all commands
/clear # Clear history
/status # System statusThe /agent command is a powerful multi-step LLM orchestrator that chains multiple AI calls to handle complex tasks:
# Fix code with systematic analysis
/agent fix the bug in main.py where the loop never terminates
# Complex refactoring
/agent refactor this monolithic code into modular components
# Detailed explanations
/agent explain how the authentication system worksPipeline stages:
- Intent Analysis - Classifies the task type
- Context Gathering - Identifies needed information
- Task Execution - Performs the main work
- Validation - Verifies output quality
- Refinement - Improves based on validation (if needed)
When to use /agent:
- Complex multi-step tasks
- Code requiring analysis and validation
- Tasks needing structured reasoning
- When quality matters more than speed
See example/agent_demo.md for detailed examples.
Note: The
/taskcommand has been deprecated. Use natural conversation instead for better results.
XandAI can intelligently create and edit files with AI assistance:
Simply ask to create a file with a specific name:
xandai> create tokens.py with authentication functions
# AI generates complete code
# System detects filename automatically
This looks like a complete python file. Save it? (y/N): y
Filename: tokens.py
File 'tokens.py' created successfully!Edit existing files by name:
xandai> edit index.py adding a health endpoint
# AI reads current file content
# Generates complete updated version
Edit file 'index.py'? (y/N): y
File 'index.py' updated successfully!The AI automatically:
- Reads files when editing (preserves existing code)
- Extracts filenames from your request
- Provides complete file content (never placeholders)
- Only prompts when you explicitly request file operations
Works with any programming language:
xandai> create app.js with Express server
xandai> edit styles.css adding dark mode
xandai> create config.json with API settingsXandAI can detect and execute code in various languages:
xandai> create a math.py that will receive two args and sum them
# AI generates complete Python script with argument handling
This looks like a complete python file. Save it? (y/N): y
Filename: math.py
File 'math.py' created successfully!
xandai> python math.py 2 2
$ python math.py 2 2
2.0 + 2.0 = 4.0
Command completed successfullyFeatures:
- Automatic code detection for Python, JavaScript, Bash, and more
- Interactive execution mode for scripts requiring input
- Non-interactive capture mode for automation
- Smart prompts for user choice between modes
AI-powered code review with Git integration. Analyzes your code changes and provides detailed feedback on security, quality, and best practices.
xandai> /review
# Automatically detects Git changes and provides comprehensive analysisAutomatically fetches and analyzes web content when you paste links:
xandai> /web on
xandai> How does this work? https://docs.python.org/tutorial
# Content is automatically fetched and analyzedgit clone https://github.com/XandAI-project/Xandai-CLI.git
cd Xandai-CLI
pip install -e .
xandaiMIT

