Skip to content

LIT Desktop (Early Access)#

LIT Desktop is the LIT Platform as a native app for your Linux, Mac, or Windows machine — free, local, and private. It runs the full chat/agent experience on your own machine: your conversations, agents, and data stay with you.

This is an early-access ("beta") build

We're shipping fast and being transparent about it. Right now the app is not yet code-signed, so macOS and Windows will show a scare screen the first time you open it — that's expected for unsigned software, and we walk you through the one-time step below. Signing and notarization are in progress (Apple approval typically takes about a business week). If you'd rather wait for the signed release, that's completely fine — this is for folks who want to follow along early.

1. Download#

Pick the installer for your operating system — these links always point at the newest build:

OS Download
Linux (Debian/Ubuntu) LIT Desktop .deb
Linux (Fedora/RHEL) LIT Desktop .rpm
macOS (Apple Silicon / M-series) LIT Desktop for Mac (Apple Silicon)
macOS (Intel) LIT Desktop for Mac (Intel)
Windows (x64) LIT Desktop for Windows

(All downloads and past versions live on the releases page.)

2. Prerequisite — connect your AI#

LIT Desktop works with Claude, ChatGPT, or Gemini. How you connect depends on whether you're using a subscription or an API key:

Install the command-line tool for your provider — the app drives it under the hood so you use your existing plan instead of paying per-token. Run it once and log in; LIT Desktop finds it automatically.

Provider Install
Claude Claude Code — then run claude once and log in
ChatGPT Codex CLI — then run codex once and log in
Gemini Gemini CLI — then run it once and log in

You only need to do this once, for whichever provider(s) you plan to use.

No CLI needed for any provider. Paste your API key into the app's setup screen (Settings → Credentials) — Anthropic, OpenAI, or Google. Usage is billed to that provider account.

3. Install & open#

  1. Open the .dmg and drag LIT Desktop to your Applications folder (either the system-wide /Applications or your personal ~/Applications — both work).
  2. Because the app isn't signed yet, clear the download quarantine once. Open Terminal, type xattr -cr (with a trailing space), then drag the app icon into the Terminal window so it fills in the exact path, and press Enter: 
    xattr -cr "/Applications/LIT Desktop.app"        # if you installed it for all users
xattr -cr ~/Applications/"LIT Desktop.app"       # if you installed it just for you
    (Pointing this at the wrong folder gives a "Permission denied" error — use the path where you actually dragged the app.)
  3. Open the app from Launchpad or Finder.
  1. Double-click the .msi to install.
  2. Windows SmartScreen will warn about an unrecognized app — click More info → Run anyway.
  3. Launch LIT Desktop from the Start menu.
# Debian / Ubuntu
sudo dpkg -i LIT-Desktop-Linux-x64.deb

# Fedora / RHEL
sudo rpm -i LIT-Desktop-Linux-x64.rpm

Then launch LIT Desktop from your app menu (or run lit-desktop).

4. First run#

  1. On the Setup screen, add a Credential — your Claude subscription login or an API key.
  2. Create an Agent (pick a model).
  3. Say hi in the #general channel. That's it — you're chatting locally.

Something went wrong?#

The app writes a full startup log to:

  • macOS/Linux: ~/.local/share/lit-desktop/logs/backend.log
  • Windows: %USERPROFILE%\.local\share\lit-desktop\logs\backend.log

Send that file to support@lit.ai and we'll take a look.

On the roadmap#

  • Connect to a server — link your desktop to a shared LIT server for team channels, sync, and always-on autonomous agents (your desktop stays free for local use; the server is where paid team features live).
  • The full deep learning platform — LIT began as a deep learning studio, and the complete ML pipeline (datasets, feature engineering, training, and prediction) is coming to LIT Desktop. Build and train models on your own machine, right alongside your agents.
  • Docking workspaces, rich Markdown & code editing, and search across your channels, calendar, and knowledge graph.