I’ve got a few ideas. Let’s talk them through.
Same agent.
New way to talk.
Your agent stays where it already works. With its tools, its context, and the task you left open.
Calls happen over the internet inside Suzu. There’s no phone number to buy for every agent.
How the connection works
Codex
Hermes
Claude Code
OpenClaw
Touch the
conversation.
See the options while you talk. Charts, comparisons, and controls appear right inside the call.
Inside interactive panelsProject week
Here’s the week. Take a look around.
A little more
like them.
A familiar agent deserves a familiar face. Pick a shape, bring an image, or give a pet a place in your day.
Make it personal
Find their voice.
Choose a voice and a supported speech model for each contact. The task agent keeps the work.
Voices and modelsOut of the chat.
Into your day.
Read the quick start- Meet on iPhone.
Suzu is being built around a hosted connection.
- Bring your agent.
Copy its setup prompt from Suzu. Your agent connects through MCP and checks its capabilities.
- Keep the conversation.
Choose its voice and appearance. Take the call, then return to your original task.
Before we call.
Is this a new agent?
Suzu connects to the agent you already use. It keeps its task context, tools, and permissions. Speech providers carry the conversation; GPT-Live can handle immediate dialogue and delegate task work to the connected agent.
Can I call an idle agent?
Agent-initiated calls work through compatible MCP harnesses. To call an idle agent from the phone, the runtime must expose a supported hook or API that resumes that exact session. Suzu assesses this separately.
Do I need to host anything?
The intended everyday experience is the iPhone app connected to hosted Suzu. Public hosted onboarding is still in development. Today’s working development setup uses one standalone service per phone; self-hosting is an advanced option.