Your First Remote Session
This tutorial walks you through setting up the PizzaPi relay server so you can stream agent sessions to a browser — and interact with them remotely from any device.
Prerequisites
Section titled “Prerequisites”Before you start, make sure you have:
- PizzaPi CLI installed — see Installation
- Docker and Docker Compose — needed to run the relay server
- An AI provider API key — Anthropic, OpenAI, Google, or a Claude Pro/Max subscription
Set Up the Relay
Section titled “Set Up the Relay”-
Start the relay server
Section titled “Start the relay server”One command builds and starts everything:
Terminal window pizzapi webWait for the success message:
✅ PizzaPi web is running at http://localhost:7492This spins up a Docker Compose stack with Redis (event buffer) and the PizzaPi server (relay API + web UI).
-
Create your account
Section titled “Create your account”Open http://localhost:7492 in your browser. You’ll see the PizzaPi web UI.
Click Sign up to create an account. This is a local account on your relay — no external service involved.
-
Connect the CLI
Section titled “Connect the CLI”In your terminal, run:
Terminal window pizzapi setupThe setup wizard will prompt for:
- Relay server URL — enter
http://localhost:7492 - Email and password — use the account you just created
┌─────────────────────────────────────────┐│ PizzaPi — first-run setup │└─────────────────────────────────────────┘Relay server URL [http://localhost:7492]: http://localhost:7492Email: you@example.comPassword: ••••••••Connecting to relay server… ✓✓ API key saved to ~/.pizzapi/config.json - Relay server URL — enter
-
Start a session
Section titled “Start a session”Navigate to any project directory and start PizzaPi:
Terminal window cd /path/to/your/projectpizzapi -
Watch it in the browser
Section titled “Watch it in the browser”Go back to http://localhost:7492. Your session appears in the session list with a live status indicator.
Click on it to see:
- Live token stream — the agent’s response appearing in real-time
- Tool calls — every file read, bash command, and edit as collapsible cards
- File diffs — syntax-highlighted diffs for code changes
- Chat input — type messages to steer the agent from the browser
What’s Next?
Section titled “What’s Next?”Now that you have a working relay, here are the natural next steps:
- Runner Daemon — Run sessions headlessly without a terminal. Spawn and manage agent sessions entirely from the web UI.
- Self-Hosting — Deploy the relay on a remote server with Docker Compose, custom domains, and production settings.
- Tailscale HTTPS — Expose your relay over a secure Tailscale tunnel with automatic TLS — access your sessions from anywhere.
- Configuration — Customize model defaults, MCP servers, hooks, and more.