App Reference

Lite Suite ships with apps organized into four categories. Every app runs locally on your machine -- no cloud accounts required.

Desktop Apps

These are standalone desktop applications with their own windows and UIs.

LiteCore

Mission control for your entire toolkit.

The central hub — installer, dashboard, embedded terminal, MCP gateway with 16 tools, Claude integration (teams, observability, hooks, profiles), process management, and the LCM memory system. Always running, always connected.

• Setup wizard with manifest-driven app selection and installation
• Dashboard with health polling, action menus, and status monitoring
• MCP gateway — centralized MCP server management with card-based UI and stats
• Claude integration — teams management, session observability, hooks configuration, and profile snapshots
• Embedded xterm.js terminal with LiteCLI panel for skill compilation
• LCM (Lossless Context Management) — DAG-based memory with automatic compaction

LiteSpeak

Voice dictation with approval gates. Nobody else has this.

Press a hotkey, speak, and LiteSpeak transcribes with LLM refinement. Unique approval gates: vocal authorization, configurable timeouts, TTS read-back, and echo prevention.

• Approval gates with vocal authorization and echo prevention
• 7 LLM providers (GPT-4.1, Claude 4.5, Gemini, local models)
• Custom modes with per-mode system prompts and model overrides
• Managed local TTS with voice selection and speed control
• HTTP API on port 7428 for programmatic access

LiteEditor

The spatial IDE for multi-agent workflows. Free and open source.

An infinite 2D canvas where editor, terminal, browser, Claude Code, and Codex live as spatial panes — not tabs. Multi-project workspaces with git worktree isolation, dev server management, and Ctrl+1-9 project switching. Free forever, no account required.

• Infinite spatial canvas — pan, zoom, and arrange panes across an unbounded 2D surface
• Claude Code and Codex VS Code extensions running natively as first-class canvas panes
• Multi-project workspaces with git worktree isolation and dev server auto-detection
• Three modes: Canvas (spatial), Zen (distraction-free), Editor (classic VS Code-style)
• Open source: github.com/ahostbr/LiteEditor

LiteTerminal

The terminal AI agents can control.

Multi-terminal with grid, splitter, tab, and window layouts. HTTP PTY Bridge API lets AI agents read, write, and control terminal sessions programmatically.

• 4 layout modes — grid, splitter, floating windows, tabs
• HTTP PTY Bridge (port 7425) with bearer token auth
• Multi-agent orchestration tested with 4 simultaneous Claude Code agents

LiteAgent

Give Claude a body. Let it run.

Autonomous AI agent with identity files (soul, user, memory), heartbeat scheduling, embedded browser, Python CLI agent loop, DAG-based memory (LCM), and local LLM support via LM Studio.

• Identity system — soul, user, and memory files for persistent self across sessions
• Heartbeat scheduling — configurable autonomous wake/work/sleep cycles
• Timeline — chronological history of every agent action
• DAG-based memory (LCM) and local LLM support

LiteDesign

AI-powered design canvas. Prompt to pixel. Free and open source.

A design canvas where AI generates, iterates, and deploys UI designs. Claude Code generates HTML/CSS and pushes it live to a browser preview panel. Model-agnostic — works with any CLI agent.

• Infinite design canvas with browser previews and Claude panels
• Design Bridge API on port 7435 — any CLI agent can render HTML and take screenshots
• Design system panel — visual editor for colors, typography, spacing
• Multi-screen projects and export/deploy to Cloudflare Pages
• Open source: github.com/ahostbr/LiteDesign

LiteImage

Local AI image generation with A/B model comparison. Nobody else does this.

Local AI image generation with Flux, SDXL, Stable Diffusion, and video models. A/B model comparison with 3 view modes, HuggingFace Hub browser, chainable pipeline, batch matrix, face swap, and video generation.

• A/B model comparison — split, overlay, and toggle views
• HuggingFace Hub browser — one-click model downloads
• Chainable pipeline and batch matrix
• Face swap with automatic landmark detection
• Video generation from text or images

LiteYT

200 videos. One click. Full analytics.

Dual-engine transcript extraction, 200-video batch processing, channel scraping hub, and analytics dashboard with 7+ chart types.

• Dual-engine extraction with automatic fallback
• 200-video batch processing for playlists, channels, or custom URL lists
• Channel scraping hub with built-in filtering
• Analytics dashboard with 7+ chart types

Frameworks

LiteTUI

TUI + CLI generation framework.

LiteTUI scaffolds production-ready Textual TUI and Click CLI applications from a single command. Describe your software, get a complete Python package with a rich terminal interface, a typed CLI, shared business logic, four themes, and Claude Code integration. Every generated project follows the shared-core architecture: business logic lives once in core/, and both the TUI and CLI import from it.

• One command generates a full TUI + CLI project
• 6 pre-built Textual components (AppShell, FeatureSidebar, StatusPanel, StreamingLog, DataExplorer, CommandPalette)
• 4 themes (dark, light, high_contrast, brand with custom accent color)
• Shared-core architecture: change the core, both interfaces stay in sync
• SoftwareBackend wraps external binaries (ffmpeg, ImageMagick, etc.)
• Session management with undo/redo history
• SKILL.md and TUI-HARNESS.md for AI agent integration
• Every generated project is pip-installable out of the box

CLI Tools

LiteCLI

Universal CLI/TUI for MCP servers + Skill Compiler. Open source.

Connect any MCP server, get CLI commands instantly. Plus a Skill Compiler that reads agent skills and generates standalone CLIs with encrypted credentials and local LLM fallback.

• MCP to CLI in seconds — JSON Schema to Click commands
• Skill Compiler — generate standalone CLIs from skill definitions
• Encrypted credentials via OS keyring
• --local flag for LM Studio routing on every compiled CLI
• Open source: github.com/ahostbr/LiteCLI

MCP Servers

These are background services that expose tools to AI agents via the Model Context Protocol. See MCP Servers for configuration details.

LiteMemory

Your AI remembers everything. Automatically.

LCM (Lossless Context Management) — a DAG-based memory system. Messages stored immutably, automatically compacted into summaries at increasing depths (minutes to hours to days to months), searchable via FTS5. Plus full Obsidian vault access.

• DAG-based memory with automatic compaction
• Multi-level summarization
• FTS5 full-text search
• Obsidian vault read/write/search

LiteMCP

Consolidated MCP server with 16 tool domains.

A single MCP server that provides filesystem, process, clipboard, screen capture, window management, system info, notification tools, and more.

• 16 domains in one server
• Filesystem operations (read, write, search)
• Process management
• Clipboard, screen, and window control
• System information and notifications

Other

LiteBench

Benchmark any LLM. Compare everything.

Desktop benchmarking app with multi-criteria scoring engine, 5 pre-built test suites (72 tests), SSE streaming, 12 chart types, and side-by-side comparison. Any OpenAI-compatible endpoint.

• Multi-criteria scoring engine with 5 preset suites (72 tests)
• 12 chart types for result visualization
• SSE streaming for real-time progress
• Support for any OpenAI-compatible endpoint

LiteAISuite

The installer and manager itself.

LiteAISuite is the installer you run to set up everything else. It handles downloading apps, creating the Python environment, managing shortcuts, and orchestrating updates. After installation, its role is absorbed by LiteCore for day-to-day management.

• App selection and installation
• Python environment setup
• Shortcut management
• Update orchestration