Connect your own model, speech and tools, and talk.
Voice Agent is a voice client for AI models you already use. It has no account, no subscription and no server of its own: you add the services, and your conversation goes only to them.
BRING YOUR OWN MODEL
Add any server with an API key or on your own network: OpenAI, Anthropic, Google Gemini, or anything OpenAI-compatible such as Ollama, LM Studio or vLLM. Choose one model for the main agent and another for a subagent.
TALK NATURALLY
Tap the microphone and speak. The app hears when you finish, sends your words and reads the answer aloud sentence by sentence. Interrupt at any time just by talking. While you are still speaking, subagents can already start looking things up, so the answer comes sooner.
YOUR SPEECH SERVICES
Use the speech recognition and voices built into your iPhone, or connect OpenAI-compatible services, whisper.cpp, Piper or ElevenLabs. Save as many services as you like, each with its own address and login, switch between them in one tap, and try each one right in Settings.
TOOLS THROUGH MCP
Connect Model Context Protocol servers over HTTP and your agent can use their tools. Turn off any single tool you do not want it to have.
YOUR IPHONE'S CALENDAR, REMINDERS AND CONTACTS
Turn on built-in tools one by one, and the agent can read and add events, manage reminders and look up contacts. Every tool is off until you turn it on.
PROFILES
Keep several agents, each with its own personality, its own set of MCP servers, and spoken or text-only replies. Switch between them in one tap, and the agent greets you to show that everything has loaded.
PRIVATE BY DESIGN
API keys stay in the iPhone Keychain. The app collects nothing and has no analytics. What you say, and what the tools you turn on read, goes only to the services you set up, under their terms.
You need an API key or your own server to use most AI models. On iPhones with Apple Intelligence you can also use Apple's on-device model in the languages it supports.
Show more
