Getting started
This guide walks you through installing Litecode and creating your first view.
1. Install Litecode
Litecode is distributed via jDeploy, which provides native installers for macOS, Windows, and Linux:
The installer pulls down a small launcher; the first run will fetch the matching JDK and the application bundle automatically. No separate Java installation is required.
2. First launch
When Litecode starts for the first time it runs a short setup wizard that:
- Detects the CLIs and tools it can integrate with (git, GitHub CLI, Docker, agent CLIs such as Claude Code or Codex CLI).
- Captures a snapshot of your environment so agent sessions launch with the
same
PATHand login environment as your shell. - Lets you point Litecode at the directory where it should keep its workspace data.
You can re-run the wizard later from the app menu, or run the CLI doctor mode to diagnose missing dependencies.
3. Create a view
Views are named workspaces. Each view groups one or more repositories and the tasks you want to work on against them.
- Choose New View from the sidebar.
- Give the view a name (e.g.
litecode). - Add one or more repositories. For each repo, pick a default branch and decide whether the view treats it as read-only.
4. Add a task
Tasks are the unit of work in Litecode.
- Click New Task inside your view.
- Give the task a title — Litecode will derive a
branchSlugfrom it that becomes the default branch name across linked repos. - Write a markdown plan. The plan supports an append-only audit trail and
can be edited by agents according to the task’s
PlanAgentPermissionsetting. - Optionally add sub-tasks, references (files or URLs), and dependencies.
5. Run an agent or open a terminal
From a task you can:
- Open a terminal session anchored to a chosen repo and branch. The terminal exposes Litecode’s MCP tools so agent CLIs can query and update the task while they work.
- Configure and run an AI agent — pick the agent type, model, schedule, and run context (temp directory, fixed directory, single repo, branch, multi-repo workspace, or task-scoped multi-repo).