Connect a Raspberry Pi

Beam runs as a CLI on Linux, so a compatible Raspberry Pi can capture agent activity without a desktop environment. Install it on the Pi where the coding agent runs, pair the Pi with your workspace, and use your laptop to view the dashboard.

The setup is independent of a particular Pi model, but runtime compatibility still matters. The current standalone installer supports Linux ARM64, not every Raspberry Pi or OS version. The npm package offers another route when a compatible Node.js runtime is available. These instructions are based on the installer and CLI implementation; they are not a physical-device certification for every Pi model.

Check the Pi and operating system

Connect using the username and hostname you configured for the Pi:

ssh YOUR_USER@YOUR_PI
uname -m
getconf LONG_BIT
cat /etc/os-release

On Raspberry Pi OS, also run dpkg --print-architecture. A kernel reporting aarch64 does not by itself establish a 64-bit userspace. Raspberry Pi documents both 32-bit and 64-bit OS images.

The npm package declares Node.js 20.17.0 or newer as its minimum; use a maintained compatible release. The Node.js 22 platform matrix includes Linux ARM64 and ARMv7, while ARMv6 is experimental. Node support varies by release and operating system. A CLI interface alone does not remove those requirements.

Install Beam on the Pi

For a compatible 64-bit installation, download the installer, inspect it, and run it:

curl -fsSL https://agentbeam.com/install.sh -o /tmp/beam-install.sh
less /tmp/beam-install.sh
bash /tmp/beam-install.sh
export PATH="$HOME/.local/bin:$PATH"
beam --version

The installer writes to ~/.local/bin. Add that PATH setting to your shell profile for future SSH sessions. An Exec format error, missing loader, or unsupported architecture message means you should recheck the OS and binary, rather than changing executable permissions repeatedly.

Alternatively, when a compatible Node.js and npm are already installed:

node --version
node -p 'process.arch'
npm --version
npm install --global @agent-beam/beam
beam --version

Use a user-owned npm installation prefix or Node version manager. Choose one installation route so your terminal and agent hooks resolve the same beam executable. A supported Beam runtime does not guarantee that every coding assistant also supports that Pi; check the assistant's own platform requirements.

Connect a headless Pi to your workspace

On the Pi, run:

beam connect

The CLI prints a connection URL and waits for authorization. Open that URL in your laptop's browser, sign in to Beam, choose the intended workspace, and confirm the Pi. Leave the SSH session running until the CLI reports success. You do not need a browser installed on the Pi for this flow.

The connection URL contains a one-time pairing token. Keep it private. If it expires, run beam connect again. An administrator-provided enrollment code is another option; see Connect a workspace.

Verify identity and fetch the current workspace policy:

beam whoami
beam account
beam sync

Connection can also attempt an entitled enterprise package installation. Inspect that result separately: successful CLI pairing does not establish that every enterprise monitoring feature supports the Pi.

Capture the first agent action

Install hooks for the supported assistants present on the Pi, then start the collector:

beam agent list
beam agent install-all
beam start

Keep this terminal open. In another SSH session, use the assistant in a disposable project and request a harmless action such as listing files. Check the workspace activity view for the Pi's device, agent, tool, and timestamp. A proposed-action event is not proof of execution or success.

Beam must run where the action occurs. Installing it only on your laptop does not expose individual commands inside a separate Pi session, and installing it on the Pi does not capture a coding agent that actually runs elsewhere. See agent hooks when no event appears.

Open local Studio through SSH

The collector binds to loopback on port 4319 by default. From your laptop, open a tunnel:

ssh -N -L 4319:127.0.0.1:4319 YOUR_USER@YOUR_PI

In an SSH session on the Pi, run beam token. Open http://127.0.0.1:4319/?token=YOUR_LOCAL_PAIRING_TOKEN in your laptop browser, replacing the placeholder with that token. Keep the tunnel running. If your laptop already uses port 4319, choose another local forwarding port and use it in the browser URL.

This provides access to the Pi's local Studio without exposing the collector to the LAN. The Studio token is different from the workspace connection token; do not share it or paste it into public reports.

Keep the collector running and add policies

After stopping a foreground collector, install the Linux user service:

beam service install
beam service status
beam service logs

The service uses systemd under your account. A Pi that must collect after logout needs its user-session lifecycle configured accordingly; ask its administrator about user lingering. Verify status and a fresh event after reconnecting or rebooting. The background service guide covers configuration and removal.

Start policy rollout with a small set of presets: recursive deletion, destructive Git commands, elevated privileges, and outbound uploads are useful candidates to review for a shared development Pi. Adding a preset leaves the mode unchanged. Save deliberately, synchronize, and verify the actual hook behavior using safe fixtures.

Include browser activity separately

A headless Pi does not run a browser extension. If you use AI websites on your laptop, install Beam's extension in that browser profile. Free Sentinel checks supported prompts locally; the separate workspace browser extension can report browser activity when its enterprise rollout is available and connected.

The Pi and browser have separate device identities. Browser submission attempts and observed responses do not establish that a command executed on the Pi. Correlate the devices and timestamps in the workspace while keeping the meaning of each record clear.