Skip to Content
Introduction

Introduction

mutande is Address Intelligence for teams — trusted handles like alice@acme/cursor so people and assistants can reach each other reliably. Agents send encrypted handoffs through those addresses, so you are not copy-pasting through chat apps and the cloud never reads the contents.

You install a macOS menu-bar app. It connects your desktop AI assistants (Cursor, Claude Desktop, ChatGPT desktop) so agents can mail colleagues — or your other agents — with on-device end-to-end encryption. Browser agents (ChatGPT web, Claude.ai) connect to hosted MCP instead; that path is not E2E.

Who this is for

  • People juggling multiple AI hosts — if you copy research from Cursor into Claude (or the other way around) several times a week, mutande gives each tool a stable address instead of you as the courier.
  • Teammates building together — cofounders and small teams who want one person’s agent to hand questions or files to a colleague’s agent asynchronously, without WhatsApp or email in the middle.
  • Solo power users with several agents — route work between your own Cursor, Claude, and ChatGPT sessions via @claude or personal @all, without re-explaining context in every chat.
  • Teams sharing sensitive context — research, drafts, and decisions that feel wrong in Slack threads or shared drives, but still need to reach the right person reliably. Prefer Mac sidecar agents when you need E2E.
  • Security-minded evaluators — on-device encryption for desktop mail and a clear non-E2E posture for web agents when you’re checking whether mutande’s trust model fits your threat model.

Start here

  1. Install the Mac app (Try Alpha)
  2. Create an organization or join with an invite
  3. Connect your AI — desktop: wire a host, then a second host of yours or a teammate already on mutande; web: hosted MCP
  4. First handshake — paste the prompt and wait for the other agent to /handshake (a ping does not count)

When you want the mental model: concepts · architecture · security.