Install FLUJO on your computer, then connect the AI and tools for a reusable Website Builder agent. Give it a site brief in chat and check the local result. Your browser is where you’ll work; FLUJO runs in the background.
The Windows setup wizard installs missing tools and builds FLUJO. It needs an internet connection and Windows App Installer (winget).
Download Windows setupUse the intended release’s installer and read its notes. Current installer source pins versioned releases; some older installers follow the development branch.
Windows setup and update details ↗Already have Node.js? Use 22.17+ within 22.x or 24.2+ within 24.x. Open a terminal and run:
FLUJOnpx flujo-ai@latestThis runs the latest published npm release, with no local build. Connected tools may still need Git, Python, uv, Bun, or their own runtime.
npm data location and update details ↗A separate clean-installation take checked the published npm package 3.46.2. It used port 4300 to keep the existing workshop session on 4200 available. The latest published version may differ when you follow the command above.
The one-line installers set up a source checkout. They follow the main development branch by default, which can be ahead of published releases. Read the script and setup notes before running it.
curl -fsSL https://raw.githubusercontent.com/mario-andreschak/FLUJO/main/scripts/install.sh | bashirm https://raw.githubusercontent.com/mario-andreschak/FLUJO/main/scripts/install.ps1 | iexDocker Compose builds the source in your checkout. Follow the Docker guide for ports, persistent data, and updates. For manual source setup, use the source installation guide.
Linux/macOS details · Corporate proxy or certificate help · Uninstalling
Open localhost:4200. If it opened automatically, use that tab. Keep the terminal or server running while you work, including for scheduled tasks.
If FLUJO is already running on that port, return to the existing instance. A new workspace can keep this website example separate from your other projects.
Open AI Setup → Connect AI. Pick your provider and follow its connection steps.
Save, then choose Test model on the connection’s card. Continue after a successful result. A saved connection alone does not prove your access works.
Provider setup help
AI access is separate. FLUJO is MIT-licensed software. It does not include a provider plan. Cloud calls—including tests and scheduled runs—may use quota or incur charges. Local models use your hardware.
Local storage, chosen connections. Your workspace is stored on your computer. Cloud providers receive the prompts and context you send; connected services receive the requests they need.
Set your privacy preferences. In Settings, choose a private encryption password for stored credentials. The public default is not a private vault. Anonymous daily usage sharing is on by default and can be changed in Privacy & Usage.
Keep the default local boundary. FLUJO’s default local profile has no owner login. Workspaces organize projects rather than separate users. Keep Network access on Localhost unless you have deliberately set up the required access controls.
Privacy, costs, and tool permissionsCheck that the terminal says FLUJO is running, then open localhost:4200. Keep the server open. If the port is already in use, find the existing FLUJO instance before starting another one.
Edit the connection and check the key or login, selected model, and available allowance. For a local model, check that Ollama is running and the model is downloaded. A login on another computer does not authenticate the server running FLUJO. Read the connection guide.
Read its installation or startup error. Check the required runtime and command on the machine running FLUJO. Test the tool directly before adding it to your assistant; don’t repeatedly reinstall it without checking the error. More recovery help.