How to Build a Raspberry Pi Information Kiosk with OpenClaw AI

A Raspberry Pi digital signage deployment in a hotel lobby logged over 1,000 consecutive days of uptime before its first scheduled maintenance, according to piSignage's 2026 optimization guide. That kind of reliability is well-documented for static display setups. But most Pi kiosk guides stop at slideshow rotation or a single web dashboard pinned in Chromium. The display shows content. It doesn't understand it.

OpenClaw is an open-source AI agent framework that runs on a Raspberry Pi and connects to cloud LLMs like Claude, GPT-4, and Gemini. Instead of displaying a fixed URL, an OpenClaw-powered kiosk can answer visitor questions, pull live data from APIs, log interactions, and adjust what's on screen based on context. The Raspberry Pi Foundation featured OpenClaw on their official blog, highlighting a build where someone created a fully functional wedding photo booth through natural language prompts alone, with the agent handling file management, hotspot configuration, and admin access without any manual coding.

The hardware cost is modest. A Raspberry Pi 5 with 8 GB of RAM runs about $95, draws between 3 and 7 watts depending on load, and costs roughly $8 per year in electricity at 24/7 operation. Add a touchscreen, case, and power supply, and you're looking at $150 to $250 total for a complete kiosk build. That's a fraction of what commercial kiosk terminals cost, and you get an AI agent that can do more than cycle through slides.

This guide covers five stages: hardware selection, OS and kiosk mode setup, OpenClaw installation, content management with persistent cloud storage, and production hardening for always-on deployment.

The component list for a Pi kiosk is short. Five items get you from zero to a working display terminal.

Raspberry Pi 5 (8 GB)

The Pi 5 is the recommended board for an OpenClaw kiosk. Its quad-core ARM Cortex-A76 processor handles the Node.js runtime, Chromium in kiosk mode, and background agent processes without thermal throttling under normal conditions. The official OpenClaw documentation lists the Pi 5 (4 GB or 8 GB) as "Best" in its compatibility matrix, with the Pi 4 (4 GB) rated "Good" and the Pi 3B+ marked as "Slow but functional." Go with 8 GB if you plan to run skills that handle concurrent API calls or process uploaded files.

Display

You have two paths here. A 7-inch capacitive touchscreen ($40 to $100 depending on brand and resolution) connects directly to the Pi's DSI port or via HDMI. The official Raspberry Pi 7-inch display is the easiest option since it needs no driver installation on Raspberry Pi OS. For larger installations, any HDMI monitor works. Lobby kiosks and wayfinding stations typically use 10 to 15-inch panels. The ClawPi project on GitHub uses a 10-inch touchscreen wired directly to the Pi 5.

Storage

Skip the SD card for a production kiosk. SD card corruption accounts for roughly 70% of Pi deployment failures in 24/7 use cases, according to piSignage's field data. An M.2 HAT+ with an NVMe SSD ($15 to $30 for the HAT, $20 to $40 for a small SSD) eliminates the most common failure mode entirely. If you stick with an SD card, use a high-endurance model rated for continuous write cycles.

Case and cooling

Active cooling keeps the Pi running between 45 and 58 degrees Celsius versus 70 to 75 degrees with passive cooling alone. The official Raspberry Pi active cooler mounts directly to the Pi 5's heatsink standoffs. For a kiosk enclosure, a vented case with the active cooler is the right combination. Sealed enclosures without airflow will thermal-throttle during sustained agent workloads.

Power supply

Use the official Raspberry Pi 27W USB-C power supply. Third-party chargers frequently cause under-voltage warnings when a display and active cooler draw power simultaneously. Voltage instability is the second most common cause of field failures after SD card corruption.

Kiosk mode locks the Pi into a single full-screen application with no desktop, no taskbar, and no way for visitors to exit to the OS. Here's how to get there from a fresh install.

Flash the OS

Use Raspberry Pi Imager to write Raspberry Pi OS Lite (64-bit) to your SSD or SD card. The Lite image skips the desktop environment, which you don't need since the kiosk runs a single browser window. Pre-configure your hostname, SSH credentials, and WiFi during the imaging step so you can connect headlessly on first boot.

OpenClaw requires a 64-bit OS. The official docs explicitly state "always use aarch64, not 32-bit ARM."

Install the display server and browser

Since the Lite image has no graphical environment, install the minimum components needed to render a fullscreen browser:

sudo apt update && sudo apt upgrade -y
sudo apt install -y labwc chromium-browser

Labwc is a lightweight Wayland compositor that the ClawPi project uses as its display server. It handles window management with minimal overhead. For X11-based setups, you can substitute Openbox, but Wayland is the default on Pi 5 and works better with its GPU driver.

Configure auto-login and auto-start

Set the Pi to auto-login to a kiosk user account (not root) on boot. Create a Wayland session file that launches Chromium in kiosk mode pointed at the OpenClaw dashboard URL. The key Chromium flags for kiosk operation are:

chromium-browser --kiosk --noerrdialogs --disable-infobars \
  --disable-session-crashed-bubble --incognito \
  http://localhost:18789

The --kiosk flag removes all browser UI. --noerrdialogs and --disable-session-crashed-bubble prevent popup dialogs that would confuse visitors. --incognito prevents session data from accumulating.

Disable screen blanking

By default, the Pi blanks the display after 10 minutes of inactivity. For a kiosk, disable DPMS and screen blanking entirely. In Raspberry Pi Configuration (or via raspi-config on the command line), navigate to the Display tab and turn off screen blanking. You can also disable DPMS at the compositor level to prevent the display from sleeping.

Hide the cursor Install unclutter to hide the mouse cursor after a few seconds of inactivity. For a touchscreen-only kiosk, you can hide it immediately on boot. Visitors tapping the screen don't need to see a mouse pointer.

Persistent cloud storage for your Pi kiosk fleet

Back every kiosk with indexed, searchable file storage. 50 GB free, no credit card, MCP-ready endpoint for your OpenClaw agent's reads and writes.

Sign Up Free

With the display and kiosk shell ready, install OpenClaw as the intelligence layer behind the screen.

Prerequisites

OpenClaw requires Node.js 22 or later on a 64-bit ARM system. Install it from NodeSource:

curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt install -y nodejs

Verify the installation with node --version. You should see v22 or higher.

Install and onboard

The official OpenClaw installer handles the download and initial setup:

curl -fsSL https://openclaw.ai/install.sh | bash

After installation, run the onboarding wizard:

openclaw onboard --install-daemon

The wizard walks you through selecting your gateway mode (local for a kiosk), configuring API keys for your preferred LLM provider, and enabling the system daemon for auto-start. Once complete, the OpenClaw gateway runs on localhost:18789, which is the URL you pointed Chromium at in the kiosk mode setup.

Verify the installation

Check that the daemon is running and enabled for boot:

openclaw status
sudo systemctl status openclaw
sudo systemctl is-enabled openclaw

If the service is active and enabled, the agent will start automatically after any reboot. Visit http://localhost:18789 in a browser (or through your kiosk Chromium instance) to confirm the dashboard loads.

Choose your LLM backend

The Pi doesn't run large language models locally. It sends prompts to cloud APIs and renders the responses. The official OpenClaw documentation recommends cloud-based models like Claude or GPT-4o-mini over local inference, noting that the Pi's CPU and memory constraints make it better suited as "an always-on controller coordinating workflows rather than processing AI responses independently." This is the right architecture for a kiosk: the Pi handles display, touch input, and agent orchestration while the LLM provider handles reasoning.

The NixOS alternative

If you prefer a declarative, reproducible system image over manual package installation, the ClawPi project provides a complete NixOS-based kiosk configuration for the Pi 5. It bundles the OpenClaw gateway, labwc compositor, Chromium in kiosk mode, and voice activation with wake words into a single Nix flake. The tradeoff is a steeper learning curve if you're not already familiar with NixOS, but the benefit is a fully reproducible system image you can reflash identically across multiple kiosk units.

A kiosk that only runs on local storage has a single point of failure. If the SD card or SSD dies, all your content configuration, interaction logs, and uploaded files go with it. For production deployments, you want kiosk content backed by cloud storage that persists independently of the hardware.

The local-only problem

Static kiosks store everything on the Pi itself. That works for a slide deck that never changes, but an AI-powered kiosk generates content dynamically. Visitor interactions, uploaded documents, generated responses, and configuration files all accumulate. Losing them means rebuilding from scratch after a hardware failure.

You could sync to S3, Google Drive, or Dropbox, but those are raw storage buckets. They hold files without understanding them. When an OpenClaw agent needs to search uploaded documents, summarize visitor feedback, or retrieve a specific configuration, it needs more than file sync.

Connecting Fast.io as the workspace layer

Fast.io provides workspaces where files are automatically indexed for semantic search and AI chat once Intelligence is enabled. For a kiosk deployment, this means the OpenClaw agent can store visitor interaction logs, uploaded documents, and generated content in a workspace that's both persistent and searchable.

The practical setup works like this: create a workspace for each kiosk deployment. Upload your content templates, reference documents, and branding assets. Enable Intelligence on the workspace so the agent can query those files through Fast.io's MCP server using Streamable HTTP. When a visitor asks a question at the kiosk, the OpenClaw agent searches the workspace's indexed files, pulls relevant context with citations, and renders the response on screen.

The free agent plan includes 50 GB of storage, 5,000 monthly credits, and 5 workspaces with no credit card required. That covers most single-location kiosk deployments. For multi-location setups, each kiosk can have its own workspace, and a central admin can monitor all of them through the same organization.

Ownership transfer for client deployments

If you're building kiosks for clients, Fast.io's ownership transfer feature lets you set up the entire workspace, populate it with content, configure the agent, and then transfer the organization to your client. You keep admin access for maintenance while the client owns the workspace. This avoids the awkward handoff of sharing credentials or rebuilding the setup from scratch in the client's account.

Syncing content updates

When you need to update what the kiosk displays or knows about, push new files to the workspace. The Intelligence layer re-indexes them automatically. The OpenClaw agent on the Pi picks up the new content through its next MCP query without any manual intervention on the device itself. This is particularly useful for kiosks in locations where you can't physically access the hardware, like a remote lobby or a trade show floor.

Getting a kiosk working on your desk is the easy part. Keeping it running for months without physical access requires deliberate hardening across storage, power, networking, and recovery.

Storage reliability

If you're running from an SD card (not recommended for production), mount /tmp, /var/tmp, and /var/log to RAM using tmpfs. This dramatically reduces write cycles to the card. Better yet, use the log2ram utility to keep system logs in memory and flush them periodically. The piSignage 24/7 guide reports that these measures, combined with high-endurance SD cards, push achievable uptime to 99.7% across fleet deployments.

The better path is booting from an NVMe SSD via the M.2 HAT+. SSDs tolerate continuous writes far better than flash memory cards, and the failure mode is more gradual (performance degradation) rather than sudden (complete corruption).

Thermal management

Active cooling is not optional for a 24/7 kiosk. The Pi 5 thermal-throttles at 85 degrees Celsius, which means the CPU slows down to protect itself. In an enclosed kiosk cabinet, ambient temperatures can push the board into throttling territory during sustained agent workloads. The active cooler fan keeps the SoC between 45 and 58 degrees Celsius under normal conditions. If your kiosk enclosure is particularly tight, consider adding a small exhaust fan to the case.

Watchdog timer

Enable the hardware watchdog timer so the Pi automatically reboots if the system hangs. The BCM2835 watchdog is built into the Pi's SoC and can be configured through systemd. If the OpenClaw daemon or the display server crashes and the system becomes unresponsive, the watchdog triggers a clean reboot within seconds rather than leaving the kiosk stuck on a frozen screen.

Network resilience

Ethernet is more reliable than WiFi for always-on deployments. If WiFi is your only option, configure the Pi to automatically reconnect after drops and set the OpenClaw agent to cache its last-known-good state so the display shows something useful even when the network is down. The agent can queue interactions locally and sync them to your cloud workspace when connectivity returns.

Remote access

Set up SSH with key-based authentication (disable password login) for remote maintenance. Tailscale or a similar mesh VPN gives you access to the Pi from anywhere without exposing SSH to the public internet. This lets you update OpenClaw, push configuration changes, and diagnose issues without visiting the physical kiosk location.

Automatic updates

Use unattended-upgrades for security patches, but pin the OpenClaw and Node.js versions to avoid breaking changes from automatic updates. Test upgrades on a staging kiosk before rolling them out to production units. For multi-kiosk deployments, a configuration management tool like Ansible can push updates to all units simultaneously.