Start an agent run on your desktop, walk away, and watch it — and approve its file edits — from your phone. Drive both coding engines remotely: the built-in SMART-Code loop and the real Claude Code CLI. Nothing to install — it's a web app. And as of v0.1.77, no project linking required.
A signed-in AgentFoundry desktop with a project open registers a live session. Your phone pairs to it — then you're in control from anywhere.
Type into the desktop's SMART-Code or Claude Code loop from your phone — it runs on your machine, where your code lives.
The model's response, tool calls, and file diffs stream to your phone live over a secure channel.
When the agent wants to write a file or run a command, an Approve / Deny card appears. Decide on the phone or the desktop — first one wins.
New in v0.1.77: your own desktop shows up on your phone with zero project linking. Just sign in with the same account on both.
The agent executes on your desktop. The phone is a remote control — nothing is copied to the cloud.
A PWA at agentfoundry.me/app. Add it to your home screen for an app-like launch.
Scan a QR code (or type a 6-character pairing code) and your phone jumps straight into the session.
Pick an engine inside the session with the SMART-Code | Claude Code toggle. Both stream back with full history; writes always ask for approval.
AgentFoundry's bundled multi-model coding agent (auto-routes Haiku / GPT-5). Send a prompt — "add input validation to the signup form and run the tests" — and watch the response, tool calls, and diffs stream in. Approve file writes from the phone or the desktop.
Flip the toggle to Claude Code and your desktop drives the real Claude Code CLI headlessly, streaming its turns back with full prompt history. Reads run unattended; when it wants to edit a file or run a command, the same Approve / Deny card appears on your phone.