Setup Guide

Get Xlaude running on your machine.

Fifteen minutes, three steps. Install Ollama (the local AI runtime), allow Xlaude to talk to it, pull a model, done. On mobile, two steps: install Xlaude as a PWA, add an API key.

Desktop

Install, configure, pull.

Works on Mac, Windows, and Linux. Pick your platform below for exact commands.

  1. Install Ollama

    Download Ollama for Mac from the official site. It installs as a menu-bar app.

    Download Ollama for macOS

  2. Allow Xlaude to talk to Ollama

    Xlaude runs in your browser and needs permission to connect to Ollama (which runs on localhost:11434). Open Terminal and run:

    # Permanently allow Xlaude's origin
launchctl setenv OLLAMA_ORIGINS "https://xlaude.xyz"

# Then restart Ollama from the menu bar

    Running Xlaude locally during development? Add http://localhost:* too — comma-separated.

  3. Pull a model

    Terminal, one command per model you want. Start with Spark (fast) and Ember (powerful):

    ollama pull llama3.2         # Spark · 3B · ~2 GB
ollama pull llama3.3         # Ember · 70B · ~40 GB (needs GPU)
ollama pull qwen2.5-coder    # Ember Code · 32B · ~20 GB
ollama pull hermes3          # Spark Rogue · 8B · unrestricted

    70B models need 48+ GB of GPU VRAM for full-speed inference. Smaller GPUs fall back to CPU and run slow — stick to 3B–13B models.

  4. Launch Xlaude

    Open the app in Chrome, Edge, or Safari 18+. It will auto-detect your Ollama instance and list every model you've pulled.

    Launch Xlaude →

  1. Install Ollama

    Download the Windows installer. It runs in the system tray and starts automatically.

    Download Ollama for Windows

  2. Allow Xlaude to talk to Ollama

    Set a system environment variable so Ollama accepts requests from Xlaude. Open PowerShell as administrator and run:

    # Set OLLAMA_ORIGINS permanently
[System.Environment]::SetEnvironmentVariable("OLLAMA_ORIGINS", "https://xlaude.xyz", "User")

# Then quit Ollama from the tray and reopen it

    Prefer the GUI? Settings → System → About → Advanced system settings → Environment Variables → New User variable: OLLAMA_ORIGINS = https://xlaude.xyz.

  3. Pull a model

    PowerShell or Command Prompt:

    ollama pull llama3.2         # Spark · 3B
ollama pull llama3.3         # Ember · 70B · needs GPU
ollama pull qwen2.5-coder    # Ember Code · 32B
ollama pull hermes3          # Spark Rogue · unrestricted

  4. Launch Xlaude

    Open the app in Edge or Chrome. It will auto-detect Ollama.

    Launch Xlaude →

  1. Install Ollama

    One-line install, works on most distros:

    curl -fsSL https://ollama.com/install.sh | sh

  2. Allow Xlaude to talk to Ollama

    Set OLLAMA_ORIGINS in the systemd service:

    sudo systemctl edit ollama.service

# In the editor that opens, add:
[Service]
Environment="OLLAMA_ORIGINS=https://xlaude.xyz"

# Then:
sudo systemctl daemon-reload
sudo systemctl restart ollama

    Running Ollama manually instead? OLLAMA_ORIGINS="https://xlaude.xyz" ollama serve.

  3. Pull a model

    ollama pull llama3.2         # Spark · 3B
ollama pull llama3.3         # Ember · 70B · needs GPU
ollama pull qwen2.5-coder    # Ember Code · 32B
ollama pull hermes3          # Spark Rogue · unrestricted

  4. Launch Xlaude

    Any Chromium browser (Chrome, Edge, Brave, Opera). Firefox doesn't ship WebGPU by default but will auto-detect Ollama fine.

    Launch Xlaude →
Zero prompts leave your machine. Once Ollama is running and the origin is allowed, every chat message routes browser → localhost:11434 → your GPU → back to the browser. Xlaude's servers never see your conversations. We never even know you ran one.
Mobile

Install the app, add an API key.

Phones can't host large models, so mobile Xlaude talks to a free API tier through your own key. The key stays in your browser only — we never see it or your prompts.

  1. Install Xlaude as an app

    iPhone: open Xlaude in Safari, tap Share → Add to Home Screen.

    Android: open Xlaude in Chrome, tap the menu → Install app.

    Either way, you get an icon, no browser chrome, and offline support.

  2. Pick a free API provider & grab a key

    OpenRouter

    Free DeepSeek · Llama

    Aggregator with multiple truly-free frontier models including DeepSeek V3 and Llama 3.3 70B.

    Get a key →

    Groq

    Free · 14,400/day

    Fastest inference on earth. Llama 3.3 70B at ~500 tokens/sec. Generous free tier.

    Get a key →

    Google AI Studio

    Free Gemini Flash

    Gemini Flash 1.5 and 2.0, free tier at 1,500 requests/day. Not open-weight — Google's model.

    Get a key →

    Our pick for most people: OpenRouter — it exposes the most free open-weight frontier models.

  3. Paste the key into Xlaude

    In the app, open Settings → API Keys, paste your key for the provider you chose, and save.

    The key is stored in your device's local storage only. We don't ever touch it.

  4. Chat

    The model picker now shows every model your key can reach. Pick one, send a message.

    Launch Xlaude →
Troubleshooting

If something doesn't connect.

  1. "Ollama not detected"

    Three things to check:

    • Is Ollama actually running? Look for the icon in your menu bar / tray.
    • Did you set OLLAMA_ORIGINS and restart Ollama? Changing the variable without restarting won't take effect.
    • Are you on the right URL? The OLLAMA_ORIGINS value must exactly match the domain you're accessing Xlaude on (https://xlaude.xyz or http://localhost:5173 for local dev).

  2. Models not showing up

    You need to pull them first. Each model is a separate ollama pull <name> command. Run ollama list in a terminal to see what's installed.

  3. Mobile key not working

    Make sure you've enabled billing on the provider (even if you'll stay in the free tier — some providers require it for API access). Also check for whitespace when pasting the key.

  4. Still stuck?

    Email support@xlaude.xyz. We reply within 24 hours.