Support & Documentation

Support & Documentation - Everything you need to get started with LMSA

#LM Studio Setup

  • Start the LM Studio server on your computer (do NOT load a model yet).
  • Enter server URL in Settings (IP:Port).
  • Load models from LMSA using the Models button in the sidebar's Options menu.
  • Start chatting!

[NOTE]

Enable "Serve on local network" in LM Studio after starting your server. Make sure your device is on the same Wi‑Fi as the server computer.

Always load models from within LMSA, not from LM Studio directly. Loading models directly in LM Studio can cause multiple models to stack in memory instead of being replaced, leading to excessive resource usage on your server computer.

Model Switching

You can switch between different models directly in the app:

  1. Open the sidebar and tap Options.
  2. Tap the Models button.
  3. Browse available models on your LM Studio server.
  4. Tap Load next to your desired model.

The app will automatically eject the current model and load the new one.

Default Model

Save a favorite model as the default so it automatically loads when you start the app:

  1. Open the Models menu from the sidebar's Options.
  2. Find your preferred model in the list.
  3. Tap the ⭐ star icon next to the model to mark it as default.
  4. Tap the star again to remove the default.

#Import LM Studio System Prompts

LMSA can import system prompts exported from LM Studio profile JSON files stored on your computer, then save them as reusable prompts in LMSA.

Where LM Studio Stores JSON

  • Windows: %USERPROFILE%/.lmstudio/config-presets
  • macOS / Linux: ~/.lmstudio/config-presets

Find the LM Studio profile .json file on your computer, then transfer it to your mobile device before importing into LMSA.

How to Import It

  1. Copy or send the LM Studio profile JSON file from your computer to your phone/tablet.
  2. In LMSA, open Options.
  3. Open Import/Export and tap Import System Prompt (LMS).
  4. When the LMSA file picker opens, choose the transferred JSON file.
  5. LMSA reads the system prompt from the profile and stores it in Options > Saved Prompts for reuse.

If the JSON contains a profile name, LMSA uses it as the prompt name; otherwise it uses the filename.

#Ollama Setup

To connect LMSA to Ollama on your local network, configure Ollama to listen on all network interfaces. Follow the OS-specific steps below.

Windows Instructions (Quick Method)

  • Open the Ollama UI, click Settings, and enable Expose Ollama to the network.

Manual configuration:

  1. Quit Ollama completely.
  2. Open Edit the system environment variables → Environment Variables.
  3. Under User variables for [YourName] click New....
    • Variable name: OLLAMA_HOST
    • Variable value: 0.0.0.0
  4. Click OK and restart your computer.

Linux Instructions

  1. Edit the Ollama systemd service:

    sudo nano /etc/systemd/system/ollama.service

  2. In the [Service] section add:

    Environment="OLLAMA_HOST=0.0.0.0"
  3. Save and exit (Ctrl+O, Enter, Ctrl+X), then restart your computer.

#OpenRouter Setup

OpenRouter is a cloud AI service that lets you access hosted models using your API key. When enabled, LMSA sends requests to OpenRouter instead of a local server.

Quick Start

  1. Create a free account at openrouter.ai and generate an API key.
  2. In LMSA Settings enable Use OpenRouter (Cloud AI).
  3. Paste your API key into OpenRouter API Key in Settings.
  4. Open the Models menu to browse cloud models.
  5. Tap Load next to a model and start chatting.

Selecting a Model

  • Open the sidebar → Options → Models.
  • The OpenRouter model catalog will load automatically.
  • Tap Load to select a model; your selection is saved for the next session.

Key Differences from Local Mode

  • No local server needed
  • Internet required
  • Usage costs may apply (per-token charges)
  • Cloud models activate instantly

[WARN]

When OpenRouter is enabled, your messages are sent to OpenRouter's servers and then forwarded to the model provider. Do not share sensitive personal information, passwords, or private data in your conversations. Your API key is stored locally on your device only.

OpenRouter also supports Zero Data Retention for certain models — see Zero Data Retention (ZDR) for details.

Troubleshooting OpenRouter

Verify your API key and re-enter it in Settings.

Ensure the OpenRouter toggle is on and a valid API key is entered.

Check your internet connection and account credits.

Turn the OpenRouter toggle off in Settings.

#Image Generation (/image)

LMSA can generate high quality AI images directly inside your chat using the /image command.

How to use it

  1. Open any chat in LMSA.
  2. Type /image followed by a description of the image you want.
  3. Send the message to start image generation.
  4. Save the generated image to your device when it finishes.

Prompt tips

For better results, describe the subject, style, lighting, camera angle, colors, background, and mood. More specific prompts usually produce stronger images.

Usage rights

Images generated with /image can be saved and used for personal or commercial purposes.

*Unlimited image generation is subject to our Fair Usage Policy.

#Voice Mode

LMSA includes a built-in, privacy-first Voice Mode that lets you have hands-free, voice-in/voice-out conversations with your AI models. Speak naturally, and LMSA will transcribe your voice, query the model (local or cloud), and read the response back to you.

How to Enter Voice Mode

You have two simple options to start a voice chat:

  • Tap the microphone icon inside the chat input field to start voice chatting within your current conversation.
  • Or, open the side menu, tap Options, and select Voice Mode from the list.

Privacy-First Design

  • On-Device Processing: Speech-to-text and text-to-speech are processed entirely on your phone. No audio data goes to third-party voice APIs.
  • 100% Private Loops: When using local models (LM Studio/Ollama), your voice never leaves your home network.
  • Local Storage: Chat histories are saved in an encrypted, on-device database.

How It Works

Once active, Voice Mode remains running for a continuous, back-and-forth conversation. The full loop works seamlessly:

Step 1

Speak naturally

No need to tap send between turns.

Step 2

Local Generation

Processed by your PC or OpenRouter.

Step 3

Listen out loud

LMSA speaks the response back.

Optimizing Voice Mode

To get the best possible hands-free experience, consider these optimizations:

  • Use smaller, faster models: Conversational speed is key for voice. A mid-sized, quantized model (7B–13B parameters) typically strikes the best balance of speed and conversational quality.
  • Custom System Prompt: Set a custom system prompt via LMSA templates instructing the model to give concise, conversational responses rather than long-winded paragraphs.
  • Keep Wi-Fi stable: Since local models communicate over your home Wi-Fi network, a stable and trusted connection ensures low latency.
  • Cloud vs Local Voice: While OpenRouter cloud models send text data to cloud servers, the speech-to-text and text-to-speech processing still remains entirely on your Android device.

For a deeper look at the privacy architecture behind real-time voice chat with local LLM models, see our Privacy-First Voice Mode page.

#Getting Started

#First Time Using LMSA?

When you first open LMSA, you'll see a Welcome Screen and an Onboarding Quick Guide that introduces you to the app. This guide helps you configure your essential preferences and get oriented.

What you need to know:

  • LMSA includes native support for English, Spanish (Español), German (Deutsch), Russian (Русский), and Simplified Chinese (简体中文)
  • You can choose your language right in the onboarding quick guide on first launch
  • You can start chatting immediately—no complex setup required!
  • The app uses AI models to generate responses to your questions
  • You can customize everything in Settings later
  • Chat is unlimited for all users — no daily caps on any tier.

#Initial Setup

  1. Select Your Language — In the onboarding quick guide on first launch, choose between English, Spanish (Español), German (Deutsch), Russian (Русский), or Simplified Chinese (简体中文). All menus, buttons, settings, system prompts, and voice mode immediately switch to your chosen language. You can change this at any time later in Settings.
  2. Allow Permissions — LMSA may ask for permission to access files and device features. These are used for:
  • Saving and exporting chats
  • Text-to-speech functionality and voice mode
  • Other features you enable in settings
  1. Unlock with Biometric (optional) — You can set up fingerprint/face unlock in Settings for quick access
  2. Choose Your First Model — The app will suggest a default model, but you can change it anytime

#The Main Interface

#Key Areas Explained

#Header Bar (Top of screen)

  • App Icon — LMSA logo
  • "+" Button — Create a new chat
  • "≡" (Hamburger Menu) — Open/close the sidebar with all navigation options

#Sidebar (Left side, toggle with ≡ menu)

The sidebar contains all navigation:

Main Menu (Options section):

  • New Chat — Start a new conversation
  • Settings — Customize app behavior (gear icon)
  • Getting Started — View onboarding guide
  • Models — Select and manage AI models (robot icon)
  • Templates — Browse and create AI personas (grid icon)
  • Saved Prompts — Access saved system prompts
  • Import/Export — Backup and restore chats
  • What's New — See latest features
  • Help — Get guidance and support
  • Premium — Upgrade or manage subscription

Active Template Indicator:

  • Shows if a persona is currently active
  • Displays template name
  • Option to disable template

Chat History:

  • Scrollable list of all your past conversations
  • Tap any chat to open it
  • Long-tap for options (delete, rename, etc.)

Bottom Links:

  • Terms of Service, Privacy Policy, About

Connection Status (Sidebar Footer):

  • Shows your current connection mode at a glance: Local, Cloud (OpenRouter), or Custom
  • Tap the status in the footer for a quick mode switch without opening full settings
  • Useful when you want to jump between local and cloud workflows during active chats

#Model Display Bar (Below header)

  • Shows the currently loaded AI model name
  • Example: "Currently Loaded: mistral"
  • Updates when you select a different model

#Chat Area (Center)

  • Display of your previous messages (usually in a light color)
  • AI responses (in a contrasting color)
  • Code blocks are highlighted for easy reading
  • Scroll up to see older messages

#Input Area (Bottom)

  • Text field where you type your question or request
  • Send button to submit your message
  • Smart reply suggestions may appear below responses (BETA feature)

#Sending Messages & Getting Responses

#How to Send a Message

  1. Tap the input field at the bottom of the screen
  2. Type your message — Ask a question, request code, get ideas, or anything else
  3. Press Enter (or tap the Send button →)
  4. Wait for response — The AI will process your request and respond

Message Tips:

  • Be specific — Detailed questions get better answers
  • Break complex requests into smaller parts
  • You can follow up with "tell me more" or related questions

#Understanding Response Features

#Code Highlighting

When the AI includes code, it appears with:

  • Color-coded syntax highlighting
  • Easy-to-read formatting
  • You can copy code blocks to use in your projects

#Regenerate Response

Don't like a response? You can regenerate it:

  1. Find the message you want to redo
  2. Tap the "Regenerate" button/icon
  3. The AI will create a new response to your original message

#Smart Replies (BETA Feature)

Quick-suggestion buttons may appear below messages:

  • Suggested follow-up conversations
  • Help you explore topics deeper
  • Just tap to ask the suggested question

#Message History

  • Your messages are automatically saved
  • Scroll up to review previous messages
  • All messages stay until you delete the conversation

#Managing Chats

#Creating a New Chat

  1. Tap the "+" button in the header (top right), OR
  2. Open the sidebar (≡) and tap "New Chat" in the Options section
  3. Start typing your first message
  4. The chat is automatically saved with a title

#Chat Titles

  • Auto-generated — The app creates a title based on your first message
  • Custom titles — You can rename chats in the sidebar (long-tap or tap menu)

#Switching Between Chats

  1. Open the Sidebar (tap ≡ if closed)
  2. Scroll through your Chat List
  3. Tap any chat to open it
  4. Your conversation history instantly appears

#Organizing Chats

  • Search chats — Use the search function in the sidebar to find conversations
  • Delete chats — Long-tap a chat name and select "Delete" (or use sidebar menu)
  • Archive — Some chat managers let you archive old conversations

#Chat Limits

  • All Users — Unlimited chats with no daily caps on any tier
  • There are no per-day limits for local models or OpenRouter on either free or premium

Tip: Save important responses as files to keep them even if you delete the chat.

#Choosing Your AI Model

#What are AI Models?

AI Models are different versions of artificial intelligence with different strengths:

  • Local Models — Run on your device or local computer (faster, no internet needed)
  • Cloud Models — Run on OpenRouter servers (more powerful, requires internet)

#Accessing Model Selection

  1. Tap the hamburger menu (≡) in the header to open the sidebar
  2. In the sidebar Options section, tap "Models" (robot icon)
  3. Or from the welcome screen, tap the "Models" button

#Available Model Types

#Local Models (Requires LM Studio or Ollama)

  • Set up a local server on your computer
  • Make it accessible to your phone
  • Models include: Llama, Mistral, Neural Chat, and others
  • Advantages: Fast, no internet, free
  • Disadvantages: Less powerful, requires setup

#Cloud Models (OpenRouter - Requires API Key)

  • Access via OpenRouter service
  • Many options: Claude, GPT, Llama variants, and more
  • Advantages: Powerful, ease of use, active development
  • Disadvantages: Requires internet, token-based pricing, daily limits

#Setting Up a Model

#Local Model Setup:

  1. Install LM Studio or Ollama on your computer
  2. Download your chosen model
  3. Start the server on your computer
  4. In LMSA Settings, enter the server address (IP:Port)
  5. Open the Models menu to load and verify the connection

#Cloud Model Setup (OpenRouter):

  1. Visit openrouter.ai and create an account
  2. Get your API key from the dashboard
  3. In LMSA, paste your API key in Settings
  4. Select your preferred model from the list
  5. Adjust settings like output length

#Switching Models Mid-Conversation

  1. Tap the hamburger menu (≡) to open sidebar
  2. Tap "Models" in the Options section
  3. Select a different model
  4. The model display bar updates to show the new selection
  5. Next response will use the new model
  6. Full conversation history is preserved

#Quick Mode Switching

You can also switch connection modes directly from the sidebar footer:

  1. Open the sidebar
  2. Tap Connection Status
  3. Choose Local, Cloud (OpenRouter), or Custom

This is the fastest way to move between server types without navigating multiple settings screens.

#Model Features

FeatureLocalCloud
SpeedVery FastFast
PowerModerateVery High
CostFreePay-per-token
Internet RequiredNoYes
Setup DifficultyMediumEasy
Model VarietyGoodExcellent

#Customizing Settings

#Accessing Settings

Method 1: Tap the hamburger menu (≡) in the header to open sidebar Method 2: In the sidebar Options section, tap "Settings" (gear icon) Method 3: From the welcome screen, tap the "Settings" button

#Settings Overview

LMSA has 20+ customizable options organized in categories:

#1. AI Behavior Settings

Temperature

  • What it does: Controls how creative vs. focused the AI is
  • Range: 0 (focused) to 1.0 (creative)
  • Example: Use 0.3 for factual questions, 0.8 for creative writing
  • Lock Option: Tap to lock your chosen temperature so it doesn't change

System Prompt

  • What it does: Instructs the AI how to behave
  • Edit it: Customize the AI's personality, tone, and style
  • Example: "You are a helpful coding assistant who explains in simple terms"

Max Tokens

  • Controls the maximum length of AI responses
  • Higher values allow longer, more detailed replies

#2. Display Settings

App Language

  • Select your preferred language: English, Spanish (Español), German (Deutsch), Russian (Русский), or Simplified Chinese (简体中文)
  • Selected in the onboarding quick guide on first launch, and changeable anytime in Settings
  • Localizes navigation menus, system dialogs, prompt templates, and voice assistant interactions

Chat Bubble Font

  • Choose your preferred font: Default, Serif, Monospace, etc.
  • Applies to chat message text only

Chat Bubble Font Size

  • 5 size options: Extra Small, Small, Medium (Default), Large, Extra Large
  • Applies to chat message text only (not menus or settings)

Auto-Scroll

  • On: New messages automatically scroll into view
  • Off: Stay focused on specific messages

Chat Scrollbar

  • Show or hide the scrollbar for navigation

#3. Advanced Features

Hide Thinking

  • For AI models that show reasoning: choose to display or hide the thinking process
  • Useful for cleaner responses vs. seeing the AI's work

Auto-Generate Titles

  • On: Chats are automatically named based on content
  • Off: Name chats manually

Biometric Unlock

  • Enable fingerprint or face unlock for app security
  • Faster than typing a password

#4. Input & Interaction

Enter Key Behavior

  • Send: Press Enter to send messages (default)
  • Newline: Press Enter to create new line; tap Send button to submit

#5. Text-to-Speech

TTS Voice Selection

  • Choose from available system voices
  • Preview voice before setting as default

#6. Other Options

Smart Replies

  • Toggle suggested follow-up questions on/off

Web Search (Optional)

  • Powered by Brave Search, providing high-quality results from the live web.
  • When enabled, LMSA pulls fresh context to improve response accuracy for time-sensitive questions.
  • Best for: current events, fast-changing tools, pricing checks, and recent releases.
  • Brave Search is built for privacy—it doesn't track you or build search profiles.
  • Keep it off for fully offline/private-only workflows.

Haptic Feedback

  • Enable/disable vibration feedback on actions

Light/Dark Theme

  • Choose visual theme preference

Saved Custom Endpoints

  • Save custom endpoint configurations in the Settings modal
  • Great for keeping multiple server profiles ready (for example: home LM Studio, remote Ollama, or other custom endpoints)
  • After saving, you can switch between saved endpoints quickly instead of retyping host, port, and related values each time

Connection Timeout

  • What it does: Sets how long LMSA waits for your local LLM server (LM Studio or Ollama) to respond before giving up and showing a connection error
  • Why adjust it: Larger models, weaker hardware, or long-context prompts can take longer to start responding — raise the timeout if you're seeing timeout errors mid-generation
  • Lower values fail fast when a server is unreachable, which is handy for quickly detecting a wrong IP/port or an offline server

#Using Personas & Templates

#What are Personas?

Personas are pre-configured AI personalities that change how the AI responds. Switch personalities instantly without retyping instructions.

#Accessing Personas

  1. Tap the hamburger menu (≡) to open the sidebar
  2. In the sidebar Options section, tap "Templates" (grid icon)
  3. Browse available personas/templates
  4. Tap one to activate it

#Built-in Personas

LMSA includes 16 pre-made templates that you can activate instantly:

  • Math Tutor — Step-by-step problem solving with explanations
  • Fortune Teller — Mystical and creative responses
  • The Joker — Humorous and entertaining
  • Story Writer — Crafts engaging narratives
  • General Tutor — Patient explanations on any topic
  • Social Media Writer — Creates engaging social posts
  • Code Assistant — Expert programming help with clean, documented code
  • Travel Planner — Itinerary and destination recommendations
  • Career Coach — Professional advice and guidance
  • Email Polisher — Refines and improves written communication
  • Sous Chef — Cooking advice and recipe suggestions
  • Debate Partner — Explores multiple perspectives critically
  • Blog Post Writer — Structures engaging long-form content
  • Tech Support — Troubleshoots technical issues systematically
  • Fitness Coach — Personalized fitness and wellness guidance
  • Summarizer — Condenses information into key points

#Creating a Custom Persona

  1. Tap the hamburger menu (≡) to open the sidebar
  2. Tap "Templates" in the Options section
  3. Tap "Create New Template" or "+" button
  4. Name it — Give it a memorable name (e.g., "My Code Helper")
  5. Set the system prompt — Write instructions for how you want the AI to behave
  6. Test it — Start a chat to see how it works
  7. Refine — Edit and improve based on results

#Custom Persona Tips

Write effective prompts:

"You are a Python expert. Provide code solutions with explanations. 
Always suggest the most Pythonic approach. Include type hints."

Include:

  • Who the AI is pretending to be
  • What tone to use
  • What topics to focus on
  • How detailed answers should be

Avoid:

  • Conflicting instructions
  • Extremely long prompts
  • Vague descriptions

#Premium Features

#What's Premium?

Premium is an optional one-time purchase ($14.99) that unlocks the complete LMSA experience with unlimited conversations, zero distractions, and full feature access forever.

#Premium vs. Free Comparison

Chat usage

Free
Unlimited
Premium
Unlimited

Voice Mode

Free
Unlimited
Premium
Unlimited

LM Studio support

Free
Premium

Ollama support

Free
Premium

File attachments

Free
Premium

OpenRouter support

Free
Premium

Web search

Free
Premium
Unlimited

Text-to-Image Generation (/image)

Free
Limited
Premium
Unlimited

Text-to-Speech (TTS)

Free
Premium

Custom endpoint usage

Free
Premium

Adjust max tokens / context length

Free
Premium

Templates & personas

Free
Premium

Save custom endpoints

Free
Premium

Offline mode

Free
Premium

PIN Unlock

Free
Premium

Biometric Unlock

Free
Premium

Advanced settings

Free
Premium

Edit system prompts

Free
Premium

Import/Export Saved Chats

Free
Premium

Ad-free experience

Free
Premium

Email Support

Free
Premium

Cost

Free
$0 (Free forever)
Premium
$14.99 one-time

#How to Get Premium

  1. Tap the hamburger menu (≡) to open sidebar
  2. In the sidebar, scroll to the "Premium" section
  3. Tap "Premium"
  4. Review the Premium Modal showing all benefits
  5. Tap "Upgrade Now"
  6. Complete the purchase flow (Google Play Billing)
  7. Instantly get unlimited access!

#Checking Your Premium Status

To quickly verify premium status, check the button on the main welcome screen and in the side menu under the premium category:

  • Premium Active means premium is activated
  • Unlock Premium means premium is not activated

#File Attachments

Attach files to your messages for AI analysis:

  • Upload documents, images, code files, or text files
  • Include them with your message to get analysis or suggestions
  • AI processes the file contents as part of your question

How to attach files:

  1. In the input area, look for attachment icon (+)
  2. Select "Upload File"
  3. Choose file from your device
  4. The file previews in the chat
  5. Type your question and send
  6. AI processes the file in your question

Supported formats:

  • Text files (.txt, .md, .csv)
  • Code files (.js, .py, .java, etc.)
  • Documents (.pdf - if supported by model)
  • Images (.jpg, .png - if supported by model)

#Export Responses (Premium Only)

Save any response as a file:

  1. Find the response you want to save
  2. Tap the "Export" or "Save" button
  3. Choose format:
  • .txt — Plain text
  • .markdown — Formatted markdown (includes code highlighting)
  1. File is saved to your device storage
  2. Access it in your file manager or open in any text editor

#Text-to-Speech

Hear responses read aloud:

  1. Tap "Speak" button on any response
  2. AI reads the message using your selected voice
  3. Choose your preferred voice model from the settings

Features:

  • Default voice available to all users
  • Multiple voice options available
  • Stops when you tap it again
  • Continues reading if interrupted

#Text-to-Image Generation

Generate AI images directly inside your chats using the `/image` command:

  1. Open any chat in LMSA.
  2. Type /image followed by a description of what you want to create (e.g., /image a futuristic cyberpunk cat wearing a glowing neon visor).
  3. Send the message.
  4. The app will generate the image using state-of-the-art diffusion models.
  5. Tap and hold or click the image to save it to your device.

Limits:

  • Free Tier: Limited to 10 total images per lifetime
  • Premium Tier: Significantly increased generation limits (subject to our Fair Usage Policy)
  • Natural Language Prompts: Use detailed English descriptions for best results
  • Commercial & Personal Rights: Saved images can be used for any project

#Offline Mode

#What's Offline Mode?

Offline Mode is a Premium-only feature that lets you run LMSA entirely on your local network without any internet connection. You can chat with AI models running on your own devices (LM Studio or Ollama) while keeping all conversations completely private and local.

#Who Can Use Offline Mode?

Premium Users Only — Offline Mode is exclusively available to users with a Lifetime Premium purchase ($14.99 one-time).

#How Does Offline Mode Work?

Setup:

  1. Install and run LM Studio or Ollama on your desktop/server
  2. Load an AI model in LM Studio or Ollama
  3. Ensure your Android device and server are on the same Wi-Fi network
  4. In LMSA, configure the server connection in Settings with your server's local IP address and port
  5. Start chatting — no internet required!

During a chat:

  • Your questions and responses stay entirely on your local network
  • Messages go directly from LMSA → your local server → LMSA
  • No data leaves your home or office network
  • Complete privacy and control

#What Works in Offline Mode?

✅ Available:

  • Chat with any local LM Studio model
  • Chat with any local Ollama model
  • All LMSA features (personas, templates, chat history, etc.)
  • Export/import chats
  • Text-to-speech (TTS)
  • File attachments
  • Custom system prompts
  • Biometric unlock

❌ NOT Available:

  • OpenRouter cloud AI models (requires internet)
  • Web Search (requires internet connection; powered by Brave Search)
  • Model downloads/updates (requires internet)
  • Any cloud-based features

#Internet Requirements

Offline Mode in Offline Mode:

  • ✅ Zero internet needed for chatting with local models
  • ✅ Your network can be completely private/firewalled
  • ✅ Perfect for sensitive conversations and private data

When Internet Is Needed:

  • Downloading new models (one-time, before going offline)
  • Updating models (optional)
  • Switching back to OpenRouter or Web Search

#Security & Privacy

Local connections are typically unencrypted:

  • Messages travel over your local Wi-Fi as plain text
  • Only use Offline Mode on trusted networks (home, office)
  • Avoid public Wi-Fi or untrusted networks
  • For remote access over the internet, consider using Tailscale (see Tailscale Setup)

#Use Cases

Offline Mode is perfect for:

  • Private conversations with sensitive information
  • Air-gapped/private networks
  • Remote work without cloud dependencies
  • Keeping your data completely local
  • Using AI without internet connectivity
  • Privacy-focused workflows

#Troubleshooting Offline Mode

"Can't connect to server"

  • Verify your server IP and port are correct
  • Ensure both devices are on the same Wi-Fi network
  • Check that LM Studio or Ollama is actually running
  • Restart the server and try again

"Slow responses"

  • Local models depend on your hardware
  • Smaller models (3B–7B) are faster
  • Larger models (13B+) use more resources
  • Close other apps to free up RAM

"Lost connection"

  • Wi-Fi may have disconnected
  • Server may have crashed
  • Try re-connecting to Wi-Fi and restarting the server

#Thinking Models & Reasoning

Some AI models (like Claude) include a "thinking" process where the AI reasons through complex problems.

How it works:

  1. When you ask a complex question, the AI shows its thinking
  2. Takes slightly longer but produces better answers
  3. You can see the reasoning process
  4. Control visibility in Settings → "Hide Thinking"

Best for:

  • Complex math or logic problems
  • Deep analysis
  • Debugging code

#LM Studio MCP (Model Context Protocol)

MCP (Model Context Protocol) allows AI models to use external tools and integrations like web lookup, weather data, code execution, and more.

#How to Set Up MCP Integrations

  1. In LMSA Settings, stay on the Local Server provider tab
  2. Tap the MCP button to open the integrations manager
  3. Add an integration via Plugin ID (references an entry in LM Studio's mcp.json, e.g., mcp/playwright)
  4. Optionally configure the Allowed Tools field for that integration
  5. Tap Done to save

Important: The Ephemeral URL option for MCP server URLs is currently disabled. LM Studio now rejects dynamic remote MCP URLs that resolve to a non-public address (this blocks LAN and Tailscale/VPN addresses). Use the Plugin ID method instead, which references a pre-configured entry in LM Studio's mcp.json file.

#Web Search (Powered by Brave Search)

Web Search lets LMSA query live web sources via the Brave Search API before the model answers. This improves timeliness and can increase answer accuracy on topics that change often while maintaining a high standard of privacy.

When to turn it on:

  1. You need up-to-date information (news, releases, prices, policies).
  2. You want the model to cross-check details with current sources.
  3. You are troubleshooting something where version updates matter.

When to keep it off:

  1. Your prompt is private/sensitive.
  2. You are working with stable evergreen topics.
  3. You prefer local-only behavior.

How to use:

  1. Open Settings.
  2. Enable Web Search.
  3. Ask your question as usual.
  4. Turn it off anytime when you no longer need live web context.

Privacy note: Web Search is opt-in and powered by Brave Search, a privacy-first search engine. When enabled, AI-generated search queries are sent to Brave to retrieve context. Brave does not track users or build search profiles, keeping your web-enhanced chats anonymous. See the Privacy Policy for more details.

#Import & Export Chats

Export all your chats:

  1. Tap the hamburger menu (≡) to open the sidebar
  2. In the sidebar Options, tap "Import/Export"
  3. Select "Export Chats"
  4. Choose format (JSON or other compatible format)
  5. Download or share the file

Import chats:

  1. Tap the hamburger menu (≡) to open the sidebar
  2. In the sidebar Options, tap "Import/Export"
  3. Select "Import Chats"
  4. Select the backup file
  5. Conversations are restored

Use case: Switch devices or backup important conversations

#Tips & Tricks

#Becoming a Power User

#Persona Stacking

Create multiple personas for different tasks:

  • "Complex Analysis" — Deep research and reasoning
  • "Quick Answers" — Direct, concise responses
  • "Creative Mode" — Writing and brainstorming
  • "Code Buddy" — Programming assistance

Tap between them as you switch tasks!

#Temperature Fine-Tuning

  • 0.2–0.3 — Factual questions, coding, math
  • 0.5–0.6 — Balanced creative/accurate writing
  • 0.8–1.0 — Creative writing, brainstorming, storytelling
  • Lock your favorite — After setting temperature in Settings, tap the lock icon so it persists across chats

#Smart Use of Models

  • Quick questions → Use faster local model
  • Complex reasoning → Use powerful cloud model (Claude, GPT)
  • Creative tasks → Try creative-focused models
  • Programming → Use code-specialized models

#Efficient Prompting

Better prompts = better answers. Examples:

❌ Weak: "Tell me about Python" ✅ Better: "Explain Python decorators with a practical example I can use in a web app"

❌ Weak: "Write code for me" ✅ Better: "Write a Python function that parses a CSV file and filters rows where age > 18, then saves to a new file"

#Multi-Step Conversations

Break complex tasks into steps:

  1. "What are the pros and cons of X?"
  2. "Now compare X and Y based on cost"
  3. "Recommend the best choice for my situation"

This guides the AI and captures nuance better than one long question.

#File Export Workflow

  1. Get perfect responses
  2. Export to .markdown files
  3. Import into notes/wiki/documentation
  4. Build your local knowledge base

#Regular Backups

  1. Periodically export your chats
  2. Save to cloud storage (Google Drive, Dropbox)
  3. Two-line protection against data loss

#Productivity Hacks

Batch Similar Tasks:

  • Ask similar questions in one chat
  • Helps model maintain context
  • Faster responses to related follow-ups

Use System Prompts Creatively:

  • "Explain results in bullet points"
  • "Assume I'm a beginner"
  • "Use simple language"
  • "Provide code examples"

Temperature Locking: Set your preferred temperature and lock it so it persists across chats. No more adjusting every time!

Custom Persona Library: Build personas for your specific needs:

  • "Email Drafting Helper"
  • "Technical Documentation Writer"
  • "Resume Reviewer"
  • "Debugging Assistant"

#Best Practices

✅ DO:

  • Start specific and refine with follow-ups
  • Use personas to shape responses
  • Export important conversations
  • Use the regenerate button to compare responses
  • Backup your chats occasionally

❌ DON'T:

  • Expect AI to be 100% accurate (always verify facts)
  • Share sensitive personal information
  • Ask AI to do anything illegal or unethical
  • Assume one answer is definitive (compare multiple models)
  • Overestimate token limits (very long messages may be cut off)

#Getting Help

#In-App Resources

  1. Help Menu — Tap "Help" in sidebar for articles and FAQs
  2. What's New — See latest features and updates
  3. Welcome Screen — Review basics anytime
  4. Contact Support — Find support info in Help menu or visit the Contact page

#External Resources

  • Official Website — Check for tutorials and blog posts

#Using Templates

Templates are pre-configured AI personas for different tasks (e.g., Math Tutor, Code Assistant).

  • Access templates from the sidebar → Templates.
  • Browse available templates and tap a template card to select it.
  • Tap Start Chatting to activate the template; a banner shows the active template.
  • Templates remain active across conversations until disabled.
  • Disable a template by clicking Disable in the template indicator banner or clearing the system prompt in Settings.

[TIP]

Larger, instruction‑following models (13B+) will adhere to templates better. For best results use a capable model.

Create and Import v2 Character Cards

LMSA templates now support full v2 character cards. You can build cards in LMSA or import existing card files from other tools.

Basic vs Advanced v2

  • Basic templates are ideal when you only need a description and a system prompt.
  • Advanced v2 card adds roleplay-oriented fields such as personality, scenario, first message, alternate greetings, tags, character book JSON, and extensions JSON.
  • LMSA stores full v2 card data and builds the runtime system prompt from that card automatically.
  • If a v2 card includes first message or alternate greetings, LMSA can auto-send the opening assistant message when activated.

How to Import v2 Cards

  1. Open Templates in LMSA.
  2. Tap Create Template and choose Advanced v2 Card for manual creation, or choose Import to load an existing card file.
  3. Select a compatible character card JSON file from your device storage.
  4. Review imported fields, then save and tap Start Chatting to activate the card.

Tip: If you edit prompts after import, save the template again so LMSA updates the generated runtime system prompt.

#Font & Text Size

Change chat bubble font style and size in Settings:

  1. Open the sidebar → Settings.
  2. Go to the Font step.
  3. Use Chat Bubble Font to choose the font style.
  4. Use Chat Bubble Font Size to set text size.

[NOTE]

Changing Chat Bubble Font Size only affects chat bubbles (messages). It does not change text size in menus, settings, or other UI.

#Language Support

LMSA includes native multilingual support across the entire mobile experience, allowing you to use the app and interact with your AI models in your preferred language.

Supported Languages

  • English (Default)
  • Spanish (Español) — Native menus, prompts, and voice responses
  • German (Deutsch) — Complete German UI and conversational chat
  • Russian (Русский) — Full Russian interface and local AI chat
  • Simplified Chinese (简体中文) — Native Simplified Chinese menus and conversations

Onboarding Quick Guide

When first launching LMSA on your device, the onboarding quick guide welcomes you and lets you choose your preferred language right away.

Selecting your language during onboarding instantly localizes the interface, navigation menus, default persona prompts, and voice mode.

Changing Language in Settings

You can switch languages at any time without resetting your data:

  1. Open the sidebar (tap the ≡ hamburger menu).
  2. Tap Settings (gear icon).
  3. Go to the Language or Display Settings section.
  4. Select your preferred language. The entire app updates immediately.

#Security & Privacy

  • All chat messages are stored locally on your device. LMSA does not store conversations on LMSA servers.
  • If OpenRouter is enabled, prompts and responses are sent to OpenRouter's servers and processed there.
  • OpenRouter traffic uses HTTPS/TLS in transit, but data is processed by third-party infrastructure.

Data Management

Clearing app cache: Safe—does not delete any data.

Clearing app storage, uninstalling, or reinstalling: Permanently deletes all data including saved chats, system prompts, templates, and settings. Exported chat files are unencrypted JSON — store them securely.

Network Security

Connections to LM Studio and Ollama servers are not encrypted. Anyone monitoring network traffic could intercept messages. OpenRouter API and chat data are fully encrypted during transit via HTTPS.

[SECURITY]

  • Only use LMSA on networks and devices you trust.
  • Never send personal information, passwords, or sensitive data through the app.
  • Avoid public Wi‑Fi or shared/untrusted networks.

Chat Export/Import

  • Export Chats to save conversations to a file (unencrypted JSON).
  • Import Chats to restore previously exported conversations; you can merge or replace existing chats.
  • Store exported files securely.

#Privacy & Premium

  • Chat messages remain private.
  • No chat data is sent to the developer.
  • Messages stay local and go straight from LMSA to your LM Studio server.

LMSA Premium: Tap Unlock Premium on the main page to upgrade to Lifetime Premium.

#Zero Data Retention (ZDR)

Look for the green shield icon

In the Models menu, any OpenRouter model marked with a green shield icon supports Zero Data Retention (ZDR). For these models, LMSA automatically tells OpenRouter to route your messages only to provider endpoints that enforce a zero data retention policy — meaning your prompts and responses are not stored by the provider.

  • No setup required on your part — LMSA automatically applies ZDR for shielded models.
  • Messages sent to a ZDR-enforced endpoint are not stored by the provider and are not used to train their models.
  • Models without the green shield icon do not support ZDR through OpenRouter.
  • If you've already configured ZDR settings directly in your OpenRouter account (globally, per model, or per request), those account-level settings always take precedence over the automatic flag LMSA sends.

LMSA is built privacy-first, so ZDR enforcement for qualifying OpenRouter models is always on and cannot be disabled within the app. Users should be aware that not all OpenRouter models listed in LMSA qualify for ZDR — those that don't will not have the green shield icon and will be treated as normal, without any special privacy routing instructions. Always check the Models menu for the green shield icon to see whether a given model is covered by ZDR.

Keep in mind that privacy-first doesn't always mean the lowest cost, fastest, or first-available option. Restricting requests to ZDR-compliant endpoints can mean fewer providers are eligible to handle a given model. During periods of high traffic, this may occasionally result in an error — if that happens, simply try again. You should also keep an eye on your OpenRouter dashboard for accurate, live usage and pricing information.

ZDR is an OpenRouter feature, not an LMSA feature — LMSA simply requests it and respects OpenRouter's compatibility list. For full details, see the Zero Data Retention section of our Privacy Policy or OpenRouter's official ZDR documentation.

#Legacy Users

Background: Early LMSA releases were paid-only. Legacy paid users can reactivate premium access in the newer app using a legacy promo code.

How to get a legacy promo code:

  1. Contact us via our contact page and include your order number and any purchase/receipt details.
  2. The team will verify and reply with a one-time promo code and instructions.

#Usage Limits

Free Users

  • ✓Chat usage: Unlimited
  • ✓Voice Mode: Unlimited
  • ✓LM Studio support
  • ✓Ollama support
  • ✓File attachments
  • ✓OpenRouter support
  • ✕Web search
  • ✓Text-to-Image Generation (/image): Limited
  • ✓Text-to-Speech (TTS)
  • ✓Custom endpoint usage
  • ✓Adjust max tokens / context length
  • ✓Templates & personas
  • ✓Save custom endpoints
  • ✕Offline mode
  • ✓PIN Unlock
  • ✕Biometric Unlock
  • ✕Advanced settings
  • ✓Edit system prompts
  • ✓Import/Export Saved Chats
  • ✕Ad-free experience
  • ✓Email Support
  • $0 (Free forever)

Premium Users

  • ✓Chat usage: Unlimited
  • ✓Voice Mode: Unlimited
  • ✓LM Studio support
  • ✓Ollama support
  • ✓File attachments
  • ✓OpenRouter support
  • ✓Web search: Unlimited
  • ✓Text-to-Image Generation (/image): Unlimited
  • ✓Text-to-Speech (TTS)
  • ✓Custom endpoint usage
  • ✓Adjust max tokens / context length
  • ✓Templates & personas
  • ✓Save custom endpoints
  • ✓Offline mode
  • ✓PIN Unlock
  • ✓Biometric Unlock
  • ✓Advanced settings
  • ✓Edit system prompts
  • ✓Import/Export Saved Chats
  • ✓Ad-free experience
  • ✓Email Support
  • $14.99 one-time

#How to Use Tailscale with LMSA for Secure Remote LLM Access

Connect to your local LLM server from anywhere, securely—without opening ports or using VPNs. This guide walks you through setting up Tailscale with LMSA, step by step. 🔒

Note: LMSA does not offer official support for using Tailscale as a secure tunnel beyond the scope of this guide.

#What is Tailscale?

Tailscale is a modern VPN service built on top of WireGuard. It creates a secure, private network between your devices without needing to open ports on your home router or deal with complex VPN configurations.

#Key Features:

  • Zero-trust security: Only devices you authorize can connect
  • Encrypted tunnel: All traffic is encrypted end-to-end
  • No port forwarding: No need to expose your LLM server to the internet
  • Easy setup: Works across different networks automatically
  • Free tier available: Perfect for personal use

#How It Works:

When you install Tailscale on your LLM server and LMSA-enabled device, they both join your private Tailscale network. Even though they're on different networks (home vs. mobile, office vs. home, etc.), they can communicate securely as if they're on the same private network.

#Why Use Tailscale with LMSA?

#📍 Access Your LLM Server Anywhere

Run a local LLM server on your home computer and access it from LMSA on your phone or tablet while traveling.

#🔒 Maximum Security

  • No open ports exposed to the internet
  • End-to-end encryption
  • Only authenticated devices can connect
  • No need to trust public cloud infrastructure

#⚡ Fast & Low Latency

Direct, encrypted connection between devices—faster than traditional VPNs.

#💰 Cost-Effective

Tailscale has a free tier perfect for home setups and small teams.

#🎯 Simpler Than Alternatives

No port forwarding, firewall rules, or complex networking knowledge required.

#Prerequisites

Before you start, you'll need:

  1. A Tailscale Account — Free account at tailscale.com
  2. Tailscale Installed on Your LLM Server — The computer running your local LLM model
  3. LMSA App — Installed on your Android phone or tablet
  4. LLM Server Running Locally — A local AI model server (e.g., Ollama, LM Studio, vLLM, etc.) accessible on your local network
  5. Network Connectivity — Both devices need internet access to connect through Tailscale

#Popular LLM Servers Compatible with LMSA:

  • Ollama — Easy, lightweight, highly popular
  • LM Studio — User-friendly GUI
  • vLLM — Fast inference server
  • LocalAI — Drop-in replacement for OpenAI API
  • GPT4All — Simple desktop app

#Step-by-Step Setup Guide

#Part 1: Install and Configure Tailscale

#Step 1A: On Your LLM Server (Desktop/Laptop)

  1. Download Tailscale
  1. Create or Sign In to Tailscale Account
  • Launch Tailscale
  • Click "Sign in" or "Log in"
  • Follow the browser redirect to sign in with:
  • Google account
  • Microsoft account
  • GitHub account
  • Apple account
  • Or create a Tailscale account
  1. Authenticate the Server
  • After signing in, you'll see a prompt to authorize this device
  • Click "Connect" or "Authorize"
  • Your LLM server is now part of your Tailscale network 🎉
  1. Note Your Server's Tailscale IP
  • In the Tailscale app, you'll see your device name and Tailscale IP (starts with 100.x.x.x)
  • Example: 100.64.45.123
  • Write this down — you'll need it in LMSA

#Step 1B: On Your Mobile Device (LMSA Phone/Tablet)

  1. Install Tailscale App
  • Search for "Tailscale" in Google Play Store
  • Download and install the official Tailscale app
  1. Sign In with Same Account
  • Open Tailscale on your phone
  • Tap "Sign in"
  • Use the same Tailscale account as your server
  • This is important—they must be in the same network!
  1. Authorize the Mobile Device
  • Follow the browser prompts to authenticate
  • Tap "Connect"
  • Your phone is now connected to your Tailscale network
  1. Keep Tailscale Running
  • Ensure Tailscale stays connected (you'll see a connected indicator)
  • On Android, Tailscale uses a VPN-like service—this keeps your secure tunnel open

#Part 2: Get Your Server's Tailscale IP

#Finding Your Server's IP Address:

On Windows/macOS:

  1. Open Tailscale app on your server
  2. Look for the main window or system tray icon
  3. You'll see your device name and IP address (format: 100.x.x.x)
  4. Click the IP to copy it to your clipboard

On Linux:

tailscale ip -4

This outputs your Tailscale IPv4 address directly.

#Verify Connectivity (Optional but Recommended):

  1. From your phone (with Tailscale connected):
  • Open a terminal/command prompt app or use ping commands
  • Try pinging your server's Tailscale IP to confirm connectivity
  • Or simply proceed to the next step—LMSA will test the connection

#Part 3: Configure LMSA Settings

Now that Tailscale is set up, configure LMSA to use your remote LLM server:

#Step 1: Open LMSA Settings

  1. Launch the LMSA app on your phone
  2. Tap the ≡ (hamburger menu) button in the top-left
  3. Tap Settings (gear icon)

#Step 2: Locate Server Configuration

  1. Look for "Server Settings" or "Connection Settings" section
  • Scroll down if needed—it may be in an "Advanced" section
  1. You should see a field for "Server URL" or "API Endpoint"

#Step 3: Enter Your Server Details

  1. Clear the existing URL (if any)
  2. Enter the new URL in this format:
   http://100.x.x.x:YOUR_PORT

Where:

  • 100.x.x.x = Your server's Tailscale IP (from Part 2)
  • YOUR_PORT = The port your LLM server is running on

Common Examples:

  • Ollama: http://100.64.45.123:11434
  • LM Studio: http://100.64.45.123:1234
  • vLLM: http://100.64.45.123:8000
  • LocalAI: http://100.64.45.123:8080

> Note: If you're not sure what port your server uses, check your LLM server's documentation or startup logs. It usually displays: "Server listening on port XXXX"

#Step 4: Save Settings

  1. Tap "Save" or "Apply"
  2. LMSA may automatically test the connection

#Testing Your Connection

#Quick Test in LMSA:

  1. Open a new chat in LMSA
  2. If your server is reachable, you should see:
  • Available models loading
  • No connection errors
  • Response times similar to local network
  1. Send a test message to confirm everything works
  2. If it works, you're done! 🎉

#If Connection Fails:

See the Troubleshooting section below.

#Troubleshooting Common Issues

Possible Causes & Solutions:

  1. Tailscale not running on server
  • ✅ Check that Tailscale is connected on your LLM server
  • ✅ Look for the Tailscale system tray icon and verify "Connected" status
  1. Tailscale not running on phone
  • ✅ Open Tailscale app on your phone
  • ✅ Ensure it shows "Connected" at the top
  • ✅ Grant necessary permissions (VPN permission required)
  1. Wrong Tailscale IP in LMSA
  • ✅ Double-check the IP address you entered in LMSA
  • ✅ Verify it matches the one in Tailscale app (format: 100.x.x.x)
  • ✅ Copy directly from Tailscale to avoid typos
  1. LLM server not running
  • ✅ Check that your LLM server is actually running on your desktop/server
  • ✅ Look for the application window or system tray indicator
  • ✅ Restart the LLM server if needed
  1. Wrong port number
  • ✅ Verify the port in LMSA matches your LLM server's port
  • ✅ Check your LLM server's documentation
  • ✅ Start logs should show: "Listening on port XXXX"

Possible Causes & Solutions:

  1. Local firewall blocking connection
  • ✅ Add exception for your LLM server port (check your LLM server's docs)
  • ✅ On Windows: Settings → Firewall → Allow app through firewall
  • ✅ Temporarily disable firewall for testing (re-enable after)
  1. Server requires API key or authentication
  • ✅ Check if your LLM server needs an API key
  • ✅ If yes, configure it in LMSA settings
  • ✅ Or generate a Tailscale auth key for additional security
  1. Devices not in same Tailscale network
  • ✅ Verify both devices are signed into same Tailscale account
  • ✅ Check Tailscale account settings at tailscale.com
  • ✅ The device should appear in "My Devices"

Possible Causes & Solutions:

  1. Internet connectivity issues
  • ✅ Tested with device on different WiFi networks
  • ✅ Check phone's internet speed (Settings → Network → Speed Test)
  • ✅ Ensure server has stable internet connection
  1. Tailscale relay being used (slower than direct connection)
  • ✅ Move devices closer together or on better networks
  • ✅ Tailscale should use direct connection automatically
  • ✅ Check in Tailscale admin panel: admin.tailscale.com
  1. LLM server overloaded
  • ✅ Check CPU/memory usage on server when running models
  • ✅ Stop other apps using the LLM server
  • ✅ Use a smaller model for better performance
  1. LMSA running too many chats or queries
  • ✅ Close unused chats in LMSA
  • ✅ Send messages one at a time (wait for response)
  • ✅ Restart LMSA if performance degrades

Possible Causes & Solutions:

  1. Tailscale IP changed
  • ✅ LLM server sometimes gets assigned a new Tailscale IP after restart
  • ✅ Check current IP again in Tailscale app
  • ✅ Update the URL in LMSA settings with the new IP
  1. Both devices not authenticated
  • ✅ Re-authenticate devices: Sign out → Sign in in Tailscale
  • ✅ Reinstall Tailscale if persistent issues
  1. Tailscale account issue
  • ✅ Log in to tailscale.com and check account status
  • ✅ Verify both devices appear in "My Devices"
  • ✅ Check for any security alerts or subscription issues

#Security Best Practices

#1. Keep Tailscale Updated

  • Regularly update Tailscale on both server and phone
  • Updates include security patches
  • Enable auto-updates if available

#2. Only Authorize Trusted Devices

  • Only add devices you own to your Tailscale network
  • Review & remove devices you no longer use at admin.tailscale.com
  • Use device approval settings for additional security

#3. Use Strong Tailscale Credentials

  • Use a strong password for your Tailscale account
  • Enable two-factor authentication (2FA) at tailscale.com/settings
  • Consider using strong accounts (Google, Microsoft) with their 2FA

#4. Secure Your LLM Server

  • Use HTTPS instead of HTTP if your LLM server supports it
  • Implement API key authentication on your LLM server
  • Keep your LLM server software updated
  • Don't expose your LLM server directly to the internet

#5. Monitor Tailscale Activity

  • Regularly check connected devices at admin.tailscale.com
  • Set up proper firewall rules on your server
  • Consider using Tailscale's advanced features (ACLs) for enterprise deployments

#6. Backup Your Tailscale Setup

  • Note your Tailscale IP scheme in case of reinstall
  • Keep device credentials secure if doing server migrations
  • Save any custom ACL rules you create

#FAQs

A: Yes! Tailscale's free tier is perfect for personal use and home projects. You get:

  • Unlimited devices (personal use)
  • Up to 3 users
  • Full end-to-end encryption
  • Direct encrypted connections Consider paying for business or team use for advanced features like SSO and device approval policies.

A: Absolutely! Install Tailscale on any device you want to connect to your LLM server:

  • Windows PC
  • Mac
  • Linux
  • iPhone
  • iPad
  • Android
  • Chromebook All devices on the same Tailscale network can access each other securely.

A: Check these common defaults:

  • Ollama: 11434
  • LM Studio: 1234
  • vLLM: 8000
  • LocalAI: 8080
  • Gradio (for many models): 7860

Otherwise:

  1. Open your LLM server app
  2. Look at the startup window/logs
  3. Search for "listening on" or "port"
  4. The number after "port" is what you need

A: This is normal. Here's how to get a permanent IP:

  1. Go to admin.tailscale.com
  2. Find your server device
  3. Click the three dots menu
  4. Look for "Disable key expiry" or "Enable stable IP" option
  5. This keeps your Tailscale IP consistent across restarts

Alternatively, just update the IP in LMSA each time (or use a hostname if using Tailscale's DNS feature).

A: Yes, but it's not recommended:

  • Tailscale acts as a VPN itself
  • Using both may cause performance issues
  • Disconnect from standard VPNs when using Tailscale
  • Tailscale is generally more secure for this use case anyway

A: Yes, Tailscale uses:

  • WireGuard encryption: Military-grade, modern encryption
  • Zero-trust security: Only authenticated devices can connect
  • Encrypted by default: All traffic is encrypted end-to-end
  • No logging: Tailscale doesn't see your traffic
  • Open source: Auditable security model

A: Yes! Tailscale works globally:

  • Your devices can be in different countries
  • As long as both have internet access, they'll connect
  • No need to change settings
  • The tunnel adapts to region changes automatically

A: Depends on your Tailscale plan:

  • Free tier: Supports up to 3 users
  • All share the same network of devices
  • Make sure you trust all family members (they'll see all connected devices)
  • For more control, upgrade to a paid plan with advanced ACLs

A: Brief outages only affect initial connection:

  • After devices connect directly, they use peer-to-peer communication
  • Direct connections don't rely on Tailscale servers
  • Tailscale strives for 99.9% uptime
  • Status page: status.tailscale.com

#Summary

You're now ready to access your local LLM server securely from anywhere! Here's what you've learned:

✅ Installed & configured Tailscale on server and phone ✅ Found your server's Tailscale IP ✅ Configured LMSA with the server URL ✅ Tested the connection successfully ✅ Understand security best practices

#Next Steps:

  1. Enjoy using LMSA with your LLM server anywhere, anytime
  2. Monitor performance and adjust as needed
  3. Explore Tailscale's advanced features as you get comfortable

#Need Help?

  • Tailscale Support: tailscale.com/contact
  • LMSA Help: Check the in-app help and settings
  • LLM Server Support: Refer to your specific server's documentation (Ollama, LM Studio, etc.)

$ lmsa connect --help

Need additional assistance?

If you ran into any issues, our support team typically replies within one business day. We're here to help!

Contact Support