lac-cli 0.3.5

AI-powered terminal toolkit with coding assistant, shell autocomplete, and API documentation generator

lac-cli

lac-cli is a terminal toolkit from lacai.io that brings AI into your command line. You get an AI shell, a coding assistant with memory, a debate engine, and an API doc generator.

What's new in v0.3.5: Added Watch Mode (/watch) - opens Chrome with voice recognition and browser monitoring, captures clicks, inputs, network requests, and sends full timeline to AI for debugging and analysis.

Install

Quick install (Unix/Linux/macOS):

curl -fsSL https://lacai.io/install.sh | bash

Quick install (Windows PowerShell):

iwr -useb https://lacai.io/install.ps1 | iex

Manual install (all platforms):

# Unix/Linux/macOS
pip3 install 'lac-cli[all]'
pip3 install anthropic

# Windows
pip install lac-cli[all]
pip install anthropic

Optional components:

pip install lac-cli[agent]  # Just the coding assistant
pip install lac-cli[shell]  # Shell with autocomplete server
pip install lac-cli[mind]   # Multi-model debate engine
pip install lac-cli[all]    # Everything

AI Provider Libraries (install based on your choice):

pip install anthropic  # For Claude
pip install openai     # For OpenAI/GPT
# Ollama requires no additional install

Getting Started

Run lac to see the menu of available tools:

lac

You'll see:

lac-cli v0.3.3

Available Tools:

  1. Shell    - AI-powered terminal with autocomplete
  2. Mind     - Multi-model debate engine
  3. GenDoc   - API documentation generator
  4. Agent    - AI coding assistant with project memory
  5. Settings - Configure AI provider and model

Select a tool [1-5] or press Enter for Shell:

Or use direct commands:

lac shell          # Launch AI terminal shell
lac mind           # Launch multi-model debate
lac gendoc <path>  # Generate API docs
lac agent          # Launch AI coding assistant

LacAgent AI Coding Assistant

Launch the AI coding assistant with project memory:

lac agent

LacAgent can:

• Read, write, and edit files
• Run shell commands with safety checks
• Test APIs with HTTP requests
• Watch files for changes
• Generate API docs
• Remember your project context
• Undo/redo changes with diffs
• Connect to MCP servers (GitHub, databases, web search)
• Monitor browser sessions with voice recognition (/watch)

Commands you can use:

/undo              # Revert last file edit with diff preview
/redo              # Reapply undone edit
/history           # Show all file edits this session
/diff <file>       # Inspect file contents with line numbers
/gendoc [path]     # Generate API documentation
/plan              # Enable PlanMode (approve plans before execution)
/noplan            # Disable PlanMode (execute immediately)
/tasks             # Show current task list
/cleartasks        # Clear all tasks
/formula list|use|off  # Manage coding formulas
/manage on|off|status  # Token management
/mcp               # List connected MCP servers
/watch [url]       # Launch browser monitoring with voice recognition

Example Usage:

[You] create a FastAPI app with user authentication
[Agent] → Creates files, installs dependencies, sets up routes

[You] run the server
[Agent] → Starts uvicorn in background, provides log file location

[You] /watch http://localhost:8000
[Agent] → Opens Chrome, monitors your testing session
        → You speak, click, type, test endpoints
        → Click "Send to AI" when done
        → AI analyzes the full timeline and suggests fixes

[You] /undo
[Agent] → Reverts last file change with diff preview

Terminal Shell

Launch the AI-powered shell:

lac shell

The first time you run it, a setup wizard will walk you through picking your AI provider and entering your API key.

To redo the setup:

lac shell --setup

To run without an internet connection:

lac shell --offline

To adjust autocomplete speed (default 150ms):

lac shell --debounce 50   # faster
lac shell --debounce 300  # slower

LacMind Multi-Model Debate

Launch the web-based debate interface for complex queries:

lac mind

LacMind runs multiple AI models in a debate format where they challenge and refine each other's ideas, then vote on the best response. Perfect for research, code generation, and complex problem-solving.

GenDoc API Documentation

Generate beautiful API documentation from your codebase:

lac gendoc /path/to/project

GenDoc scans your project, detects the framework, and uses AI to analyze your routes and controllers to generate comprehensive API documentation. Supports Laravel, Django, FastAPI, Flask, Express, and Rails.

Optional flags:

lac gendoc /path/to/project --prompt "Focus on authentication endpoints"
lac gendoc /path/to/project --integrate  # Add /docs route to your framework
lac gendoc /path/to/project --output custom.html

MCP (Model Context Protocol) Integration

Connect your agent to external MCP servers for extended capabilities:

lac mcp add    # Add a new MCP server
lac mcp list   # List configured servers
lac agent      # Start agent with MCP tools

Popular MCP Servers:

• GitHub - Create issues, PRs, search code
• PostgreSQL - Query databases, inspect schemas
• Filesystem - Safe file access to specific directories
• Brave Search - Web search capabilities
• Google Drive - Access Drive files
• Slack - Send messages, manage channels

Example Usage:

[You] create a GitHub issue about the bug we found

[Agent]
  ⏺ github__create_issue(...)
  ✓ Issue #42 created

Created issue #42 in your repository!

MCP servers are automatically connected when the agent starts. See MCP_QUICKSTART.md for setup guide.

Formula System

Create custom coding rules that get injected into the AI's instructions.

lac formula add    # Create a new formula
lac formula list   # See all formulas
lac formula use    # Activate one
lac formula off    # Turn it off

You can also use /formula inside the agent.

Token Management

Control how much context the AI uses. Helps manage costs and keeps conversations focused.

lac manage on       # Turn on token tracking
lac manage status   # See current usage
lac manage off      # Turn it off

Three modes:

• trim keeps only recent messages
• warn just shows warnings
• compact has AI summarize old messages

Use /manage inside the agent to check status.

How It Works

Main Menu

When you run lac without arguments, you get an interactive menu to choose between:

• Shell - AI-powered terminal with real-time autocomplete
• Mind - Multi-model debate for complex queries
• GenDoc - API documentation generator
• Agent - AI coding assistant with project memory
• Settings - Configure AI provider and model

You can also use direct commands like lac shell, lac mind, lac gendoc <path>, or lac agent.

LacAgent Coding Assistant

LacAgent runs in your project folder and can access files and run commands.

What it does:

• File stuff: read, write, patch files (no full rewrites)
• Glob patterns: find files like **/*.py or src/**
• Commands: runs shell commands, detects long ones and backgrounds them
• API testing: makes HTTP requests
• File watching: monitors files for changes
• Docs: generates API documentation
• Memory: remembers your project across sessions
• Undo/redo: full edit history with diffs
• PlanMode: shows you what it'll do before doing it (on by default)
• Tasks: tracks multi-step work
• Questions: can ask you stuff when it needs to
• MCP: connects to external servers for more tools

Smart stuff:

• Runs things like npm start in background automatically
• Saves logs to ~/.lac/logs/
• Paste multiple lines no problem
• Type / to see all commands
• Security checks on dangerous commands

How editing works:

1. Reads the file first
2. Makes surgical edits with patch_file
3. Snapshots before every change
4. You can /undo anytime

Terminal Shell

When you run lac shell, it starts a server in the background. You don't need to do anything.

As you type, it shows ghost text suggestions. Press Tab to use them. Type plain English like "show all files bigger than 100mb" and it converts it to the actual command, then asks if you want to run it.

The shell remembers your recent commands and their output, so suggestions get smarter as you use it.

LacMind Debate Engine

LacMind makes multiple AI models debate each other:

1. Models discuss and challenge ideas
2. They vote on who had the best reasoning
3. Winner gives the final answer

Models are anonymous (Model A, B, C) so they can't be biased. You pick how long debates run and which models join.

GenDoc Documentation Generator

GenDoc makes API docs automatically:

1. Scans your project and finds your framework
2. Sends route/controller files to lacai.io for analysis
3. Creates interactive HTML docs
4. Can add a /docs route to your app

First time you run it, you'll need an API key from lacai.io/dashboard/keys. The docs include endpoints, parameters, examples, and you can export to Swagger or Postman.

Supported Providers

Provider	Notes
claude	Anthropic API
openai	OpenAI API
ollama	Local models, no API key needed
custom	Any OpenAI compatible endpoint

Commands

Main Menu

Command	What it does
lac	Show interactive menu
lac shell	Launch AI terminal shell
lac mind	Launch LacMind debate UI
lac gendoc <path>	Generate API documentation
lac agent	Launch AI coding assistant
lac mcp add	Add MCP server interactively
lac mcp list	List configured MCP servers
lac mcp remove	Remove an MCP server

LacAgent

Command	What it does
lac agent	Launch AI coding assistant
/undo	Revert last file edit with diff
/redo	Reapply undone edit
/history	Show all file edits this session
/diff <file>	Inspect file with line numbers
/gendoc [path]	Generate API documentation
/plan	Enable PlanMode
/noplan	Disable PlanMode
/tasks	Show current task list
/cleartasks	Clear all tasks
/formula list|use|off	Manage coding formulas
/manage on|off|status	Token management
/mcp	List connected MCP servers
/watch [url]	Launch browser monitoring with voice
Ctrl+C	Interrupt agent execution
exit	Quit the agent

Terminal Shell

Command	What it does
lac shell	Launch the AI shell
lac shell --setup	Re-run setup wizard
lac shell --offline	Run without server
lac shell --debounce MS	Set autocomplete delay
exit	Quit the shell
logout	Delete your config and start fresh
clear	Clear the screen and session history
cd <path>	Change directory

LacMind

Command	What it does
lac mind	Launch LacMind web interface
Settings page	Add/edit/delete AI models
Duration selector	Set debate time (30s - 5min)
Stop button	End debate early
Export	Save conversations to PDF

GenDoc

Command	What it does
lac gendoc <path>	Generate API docs from project
--prompt "text"	Custom instructions for AI analysis
--integrate	Add /docs route to framework
--output file.html	Custom output filename
Export buttons	Download as Swagger or Postman

Config

Your config is saved at ~/.lac/config.json after setup. You can edit it directly if needed.

{
  "provider": "claude",
  "api_key": "sk-...",
  "model": "claude-haiku-4-5-20251001",
  "base_url": "https://api.anthropic.com",
  "server": "ws://localhost:8765"
}

Features

LacAgent Coding Assistant

File stuff:

• Read, write, patch files (exact string matching)
• Glob patterns like **/*.py, *.js, src/**
• Undo/redo with full history and diffs

Commands:

• Run shell commands with safety checks
• Auto backgrounds long commands (flutter run, npm start)
• Logs saved to ~/.lac/logs/

Browser Monitoring:

• /watch opens Chrome with voice recognition
• Captures clicks, inputs, navigation, network requests
• Sends full timeline to AI for debugging
• Lazy installs playwright when first used

Other:

• HTTP requests for API testing
• File watching for changes
• API doc generation
• Project memory (remembers context)
• PlanMode (approve before execution)
• Task tracking with status icons
• Ask questions when needed
• Multi-line paste support
• Slash command autocomplete
• Ctrl+C to stop anytime
• Security validation
• MCP server integration

Terminal Shell

• Ghost text autocomplete as you type
• Plain English to command conversion
• Session history tracking for better suggestions
• Adjustable autocomplete speed
• Works with Claude, OpenAI, Ollama, or custom endpoints
• Offline mode with history fallback
• Auto starts server in background
• Logout clears everything

LacMind

• Multiple models debate sequentially
• Anonymous voting for best reasoning
• Real-time streaming
• Chat history saved
• Add/edit/delete models
• Set debate duration
• Stop early if consensus reached
• Export to PDF
• Dark theme UI

GenDoc

• Auto detects framework (Laravel, Django, FastAPI, Flask, Express, Rails)
• AI analyzes endpoints
• Interactive HTML docs with search
• Resizable sidebar with filtering
• Try It feature to test endpoints
• Export to Swagger or Postman
• Optional /docs route integration
• Custom prompts for analysis
• Separate API key system
• Credit based usage

About

lac-cli is from lacai.io. Made for developers who work in the terminal.

License

MIT