Get started
Home › Getting started
Getting started

From download to a connected AI in a few minutes

Four steps: install Codelux AI, sign in, add your project, connect your AI tool. If you prefer, your AI can do most of the last step for you.

Download and start Codelux AI

Get the Free version from the download page (Windows, macOS, Linux). On Multi or Team, download your personal version from the Console, under My account › Downloads. Extract the ZIP into a folder of your choice.

Recommended: install it as a background service, so it starts with your computer and is always there for your AI. From the folder you extracted:

Windows, in a terminal opened as Administrator:

codelux.exe install-service

Linux (it may ask for your password):

./codelux install-service

Just trying it, or on macOS: start it in a terminal instead (codelux.exe start on Windows):

./codelux start

Your system may warn you the first time, because the app is not code-signed yet: on Windows choose More info › Run anyway; on macOS right-click the app and choose Open. When it starts, it shows the address of its dashboard (normally http://localhost:8766): open it in your browser.

Sign in, or start Free

In the app, create your free account or sign in. Free needs no credit card and has no time limit: one active project per computer, every feature. Multi and Team unlock more projects and your colleagues, on the same account.

Add your project

In the app, open Projects, click + Add project and choose the folder of the project you work on. Codelux AI prepares it by itself; a large project can take a little while. On Free, adding a second project means switching the active one.

Connect your AI tool

Pick yours below. You can connect several: they will share the same project memory.

Let your AI do it. In Claude Code, after installing the plugin, just ask: “Set up Codelux AI for this project.” It checks that Codelux AI is running, that your project is added, and that everything answers. With any other tool, ask your AI to read the official guide at codelux.ai/help and connect itself; it will tell you if something is missing. It never needs your password.

Setup for each AI tool

What each one supports is on Works with. If you changed Codelux AI's address in its settings, use yours instead of localhost:8766.

Claude Code Full support

  1. In a terminal, install the plugin: 
    claude plugin marketplace add codeluxai/codelux
claude plugin install codelux@codelux
    or, inside Claude Code, type /plugin marketplace add codeluxai/codelux and then /plugin install codelux@codelux.
  2. Restart Claude Code in your project folder. From now on the project memory is loaded at every session start.
  3. Type /codelux:status to check it is connected.

On Windows the plugin needs Git Bash installed. If Codelux AI is not running, the plugin simply stays quiet.

Codex Experimental

  1. Add Codelux AI to Codex's MCP servers with this address (Codex's documentation says where): 
    http://localhost:8766/mcp
  2. For the automatic memory and the edit guard: get the plugin folder from github.com/codeluxai/codelux, copy its hooks/codex.hooks.json to .codex/hooks.json in your project (or in your home folder for every project), and replace CODELUX_PLUGIN inside it with the full path of the plugin folder.
  3. Restart Codex. It also reads the Codelux AI section of your project's AGENTS.md.

On Windows this needs Git Bash. Some details of Codex's configuration are not yet confirmed: if one does not match, that step does nothing and never blocks you.

Gemini CLI Experimental

  1. Add Codelux AI to Gemini CLI's MCP servers with this address (Gemini CLI's documentation says where): 
    http://localhost:8766/mcp
  2. For the automatic memory and the edit guard: from the plugin folder (github.com/codeluxai/codelux), merge the hooks part of hooks/gemini.settings.json into .gemini/settings.json in your project, replacing CODELUX_PLUGIN with the full path of the plugin folder.
  3. Gemini reads GEMINI.md: in its settings, set contextFileName to AGENTS.md so it also reads the Codelux AI section.

On Windows this needs Git Bash. Some details are not yet confirmed by Gemini CLI's documentation.

Cursor Connected

  1. Create or open .cursor/mcp.json in your project (or ~/.cursor/mcp.json for every project) and add Codelux AI next to any servers already there: 
    {
  "mcpServers": {
    "codelux": { "url": "http://localhost:8766/mcp" }
  }
}
  2. Restart Cursor, or reload its MCP servers.
  3. Optional: add a Cursor rule asking the agent to read the Codelux AI memory when it starts and save it when it finishes.

VS Code Connected

  1. Create or open .vscode/mcp.json in your workspace and add: 
    {
  "servers": {
    "codelux": { "type": "http", "url": "http://localhost:8766/mcp" }
  }
}
  2. Reload VS Code; the tools appear in the chat's agent mode.

Windsurf Connected

  1. Open ~/.codeium/windsurf/mcp_config.json and add: 
    {
  "mcpServers": {
    "codelux": { "serverUrl": "http://localhost:8766/mcp" }
  }
}
  2. Restart Windsurf.

JetBrains Connected

  1. Settings › Tools › AI Assistant › Model Context Protocol › Add server.
  2. Use this address: 
    http://localhost:8766/mcp

Zed Connected

  1. Open Zed's settings and, under context_servers, add Codelux AI with this address: 
    http://localhost:8766/mcp

Any other tool that supports MCP Connected

  1. In your tool's MCP settings (see its documentation), add a server with this address: 
    http://localhost:8766/mcp
  2. Codelux AI must run on the same computer as the tool.

Check that it works

Ask your AI: “What do you remember about this project from Codelux AI?” On a new project it answers that there is no memory yet — that is fine. At the end of the first task, ask it to save its notes; the next session will start from them.

  • The tools do not show up? Codelux AI was probably not running when you opened your AI tool: start it, then restart the tool.
  • “This folder is not indexed”? Add the project folder in the app (step 3).
  • Still stuck? The official guide answers most questions, or write to us.

Ready?

Free: one project, every feature, no time limit, no credit card.

Download Free Works with