Cipher AI
by LMBridge

AI that's truly
private & yours.

Cipher AI runs powerful language models directly on your iPhone. No cloud. No subscriptions. No one reading your conversations.

Download Free on the App Store Connect your own models →
On-device inference
No account required
Works offline
Free
iOS 18+
Features

Built different.

Most AI apps send everything to the cloud. Cipher doesn't. Your conversations stay on your device, processed by powerful models running entirely locally.

Runs entirely on your iPhone

Cipher runs powerful language models directly on your device. No cloud, no servers, no middlemen — just fast, private AI that's always in your control.

  • Runs powerful language models entirely on-device
  • Full offline support after model download
  • Reasoning / thinking model support
  • Token streaming — no waiting for full responses
  • No account, no login, no subscription
Model Picker
📱Mistral 7B Q4_K_MOn-device
📱Phi-3 Mini Q8On-device
🌐DeepSeek-R1 70BLMBridge
🌐Llama 3.3 70B InstructLMBridge
Switch between local and remote models instantly.

Upgrade to your own big models

When you need more horsepower, connect Cipher to large models running on your Mac, PC, or Linux machine via LMBridge. Frontier-scale inference without paying per token.

  • Stream any model from LM Studio or Ollama
  • Pair with a simple 6-digit PIN — no config
  • Auto-reconnects when you reopen the app
  • Switch local ↔ remote mid-conversation
  • Model weights never leave your machine
Connected — Chris's Mac Studio
DeepSeek-R1 70B · streaming
via LM Studio / Ollama
LMBridge Relay
End-to-end WebSocket · wss://lmbridge.org
🔒 Weights stay on your machine. Only tokens cross the wire.

Models that can actually do things

Go beyond chat. Cipher exposes a growing catalog of MCP (Model Context Protocol) servers — giving your AI the ability to run commands, query databases, browse the web, and more.

  • 40+ integrations in the built-in MCP store
  • Terminal, Docker, Kubernetes, databases, web
  • Per-tool approval before anything runs
  • Auto-approve mode for trusted workflows
  • Add your own custom MCP servers
⚙️ Tool Request
terminal · run_command
docker ps --format table
Lists all running containers on your server.
Deny
Approve
You approve every tool call. Nothing runs without your tap.
Built-in Tools

Useful right out of the box.

No setup needed. These tools are available in every conversation, running entirely on your device.

🔍
Web Search
DuckDuckGo — private, no tracking
🌤️
Weather
Current conditions anywhere
💱
Currency
Live exchange rates
📐
Unit Converter
Length, weight, temp & more
🧮
Calculator
Math expression evaluator
🔑
Password Gen
Strong, random passwords
📖
Wikipedia
Instant article summaries
🕐
Timezone
Look up any timezone
🧠

Thinking Models

See the model reason step-by-step before answering. Full support for DeepSeek-R1 and other reasoning models.

Token Streaming

Responses appear word-by-word as the model generates them. No spinner, no waiting.

📶

Fully Offline

Download a model once and it works forever — airplane mode, no signal, no problem.

🎛️

Full Control

Tune temperature, context length, system prompts, and max tokens per conversation.

🔄

Auto-Reconnect

Background the app, come back later — Cipher picks up the LMBridge connection right where you left off.

🛡️

Zero Telemetry

No usage analytics, no conversation logging. The developer literally cannot read your chats.

LMBridge

Your Mac's brain,
in your pocket.

LMBridge is a lightweight relay app you install on your computer. It connects your local LM Studio or Ollama models to Cipher over a secure WebSocket — so you can run a 70B model from your couch.

🤖
LM Studio / Ollama
Your computer · any model
💻
LMBridge Desktop
Mac · Windows · Linux
🌉
Relay
wss://lmbridge.org
📱
Cipher
iPhone · anywhere

Pair Cipher to your computer with a 6-digit PIN. Model weights never leave your machine — only tokens are relayed.

Download LMBridge

Get LMBridge for your computer.

Install LMBridge on the machine running LM Studio or Ollama, then pair it with Cipher on your iPhone using a 6-digit PIN. Available for macOS, Windows, and Linux.

🍎
Apple Silicon
M1 / M2 / M3 / M4 and later
↓ Download .dmg (arm64)
v1.1.0 · .dmg installer · macOS 12+
💻
Intel Mac
2019 and earlier Intel Macs
↓ Download .dmg (x64)
v1.1.0 · .dmg installer · macOS 12+
📋 Installation Guide — macOS
1
Download & open the .dmg

Double-click the downloaded file. Drag Cipher LMBridge into your Applications folder.

2
Allow the app to open

macOS may show a security warning on first launch. Go to System Settings → Privacy & Security and click Open Anyway. This only happens once.

3
Start LM Studio or Ollama

Open LM Studio and enable the local server (default port: 1234), or start Ollama which runs on port 11434. Load the model you want to use.

4
Launch Cipher LMBridge

Open the app from your Applications folder. It connects to LM Studio or Ollama automatically and displays a 6-digit PIN in the menu bar.

5
Pair with Cipher on your iPhone

In Cipher, go to Settings → LMBridge → Add Server. Enter the PIN shown on your Mac. You're connected — your Mac's models will appear in the model picker.

🪟
Windows x64
Most Windows PCs (Intel / AMD)
↓ Download Installer (.exe)
v1.1.0 · NSIS installer · Windows 10+
🪟
Windows ARM64
Surface Pro X · Snapdragon PCs
↓ Download Portable (.zip)
v1.1.0 · portable .zip · Windows 11+
📋 Installation Guide — Windows
1
Run the installer (x64) or extract the zip (ARM64)

x64: Double-click the .exe and follow the setup wizard. LMBridge installs to Program Files and adds a Start Menu shortcut.

ARM64: Extract the .zip and run Cipher LMBridge.exe directly — no installation needed.

2
Allow through Windows Defender SmartScreen

You may see a SmartScreen prompt on first run. Click More info → Run anyway. The app is safe — SmartScreen warns about any unsigned new app.

3
Start LM Studio or Ollama

Open LM Studio and enable the local server (port 1234), or run Ollama (port 11434). Load your model of choice.

4
Launch Cipher LMBridge

Start the app. It appears in your system tray and displays a 6-digit PIN to share with Cipher on your iPhone.

5
Pair with Cipher on your iPhone

In Cipher, go to Settings → LMBridge → Add Server and enter the PIN. Your Windows models will appear instantly in the model picker.

Paste this in your terminal. It detects your architecture and installs automatically.
$ curl -fsSL https://lmbridge.org/install.sh | bash
Supports x64 and ARM64 · Ubuntu · Debian · Fedora · Arch · and more
What the script does
1
Detects your architecture

Automatically picks the right build — x64 or ARM64.

2
Downloads & installs LMBridge

Places the app in ~/.local/bin/ and makes it executable.

3
Run it, then pair with Cipher

Launch LMBridge, start LM Studio (port 1234) or Ollama (port 11434), and enter the displayed PIN in Cipher under Settings → LMBridge → Add Server.

ℹ️ Prerequisite: LMBridge works with LM Studio (port 1234) and Ollama (port 11434). Either must be running on the same machine before you launch LMBridge.

Private. Powerful.
Yours.

Download Cipher AI free and start chatting privately today. No account. No cloud. No catch.

Download on the App Store Get LMBridge for your computer