Quickstart

This page takes you from nothing installed to a first task finished by the agent. There are three moves: install Lume and sign in, open a project, then describe what you want and review the diffs that come back. Everything after that is a variation on the third step.

Supported operating systems

Lume is currently distributed for Windows.

Operating system	Architecture	Version
Windows	x64	Windows 10, Windows 11

Note: Builds for macOS (Apple Silicon) and Linux (.deb / .rpm, x64) are listed on the download page but are not yet available to download. The table above is the honest current state, not the plan.

Install Lume and complete initial setup

Download the installer from the download page and run it.

Lume installs as a normal desktop application and opens on an empty window.
Sign in. Click Log in in the chat panel.

Your browser opens on lume.computer. Sign in there, and the app picks the session up on its own — you never paste a token back into the editor.

If you are coming from another editor, Lume can import your extensions and configuration from VS Code, Cursor or Windsurf, so your keybindings and language tooling are there before you start.

Note: Signing in is what connects the editor to your account balance. The editor and code completion work without it; AI chat and the agent need an account with paid balance. See Plans and billing.

Open a project

The agent works inside a folder, so open one before you start a conversation.

Choose File → Open Folder.
Pick the root of your project — the folder containing your .git directory, not a subfolder.

The file tree appears in the sidebar and the agent's working directory is set to that folder.

Use IDE or SOLO mode

Lume has two layouts over the same engine, and Ctrl+Alt+\ switches between them at any time.

IDE mode puts the editor first with the agent in a side panel. Use it when you are writing code yourself and want help alongside.
SOLO mode puts the task first, with its own tool panels around the conversation — Doc, Terminal, Preview, Diff, History, and integrations. Use it when you want to state a goal and review the finished work.

There is no wrong choice here and no migration between them — it is the same project, the same conversation history, and the same agent.

Collaborate with the agent

Open the chat panel, type what you want in plain language, and send it. The agent reads the files it needs, proposes edits, and runs commands when the task requires it. Every file change arrives as a diff you accept or reject, and every command it wants to run is governed by the auto-run mode you have set — by default it asks first.

Two habits make the difference early on. Point at the specific file with # instead of describing it, and say what "done" looks like — a passing test, a page that renders, a build that completes. The agent can verify its own work when it knows what to verify against. Chat covers this in full.

Turn on code completion: CUE

CUE is Lume's inline completion. It writes the rest of the line or block as you type, and you accept it with Tab. Its status bar entry sits at the bottom right of the window: click it to toggle CUE on or off, review the accept and reject shortcuts, or snooze it for a set period when you want to type undisturbed.

CUE is separate from the agent and separate from your balance — it does not consume AI usage.

Use MCP servers

Model Context Protocol servers give the agent tools beyond the ones built in: querying a database, calling an internal API, driving another application. Lume can connect to MCP servers and expose their tools to the agent alongside its own, and you configure them from Settings → MCP. A dedicated page on adding and scoping servers is coming with the next section of these docs.

What is Lume?

Chat