Nevo Pi is a dedicated AI agent appliance -- a Raspberry Pi 5 pre-configured with Nevo that runs as a headless, always-on service on your local network, available as a $200 hardware product from nevo.systems. Plug it in, connect it to your network, add your API key, and you have a self-improving AI agent running 24/7 without tying up your personal computer.
The concept is simple: an AI agent should work like a utility. You do not think about your router -- it just runs. Nevo Pi brings that same philosophy to AI. It sits on your desk or in a closet, handles the tasks you send it, gets smarter over time, and never interrupts what you are doing on your main machine.
This guide walks you through everything from unboxing to running your first task. No prior experience with Raspberry Pi or Linux is required. If you can plug in an Ethernet cable and open a web browser, you can set up Nevo Pi.
For background on what Nevo is and how it works, see What Are AI Agents? and Autonomous AI Agents.
What Is Nevo Pi?
Most AI tools live on your laptop as apps competing for your CPU, your RAM, and your attention. When an AI agent is running a complex multi-step task, your machine slows down. Your fans spin up. Your battery drains. And if you close the lid, the work stops.
Nevo Pi solves this by moving the agent off your machine entirely.
Nevo Pi is a self-contained AI agent device that connects to your home or office network and operates independently of your personal computer. It runs the full Nevo stack -- OpenClaw daemon, agent orchestration, memory pipeline, QMD search, quality pipeline, and self-improvement engine -- on a Raspberry Pi 5 with an NVMe SSD for fast storage.
The reasoning happens in the cloud via Anthropic's Claude API. What runs locally on the Pi is everything else: task orchestration, agent coordination, memory management, file operations, tool execution, and the web dashboard you use to interact with it. This division of labor means the Pi handles the workload that does not require a frontier language model, while the language model handles the parts that require intelligence.
The result is an AI agent that runs continuously, costs pennies per day in electricity, and never competes with your other work.
What Is in the Box
Nevo Pi ships as a complete kit. You do not need to source components or assemble anything.
| Component | Specification |
|---|---|
| Raspberry Pi 5 | 8 GB RAM, Broadcom BCM2712 quad-core Cortex-A76 at 2.4 GHz |
| NVMe SSD | 256 GB M.2 NVMe, connected via Pi 5 M.2 HAT |
| Case | Aluminum passive cooling enclosure with M.2 slot integrated |
| Power supply | Official Raspberry Pi 27W USB-C power supply |
| Ethernet cable | 1 meter Cat6 cable |
| Pre-installed software | Nevo OS (Raspberry Pi OS Lite 64-bit + full Nevo stack) |
Why Raspberry Pi 5?
The Pi 5 is not a toy. Its quad-core Cortex-A76 handles orchestration, coordination, and memory pipeline operations easily. The 8 GB of RAM is sufficient because heavy reasoning happens on Anthropic's servers, not locally. The NVMe SSD is the key upgrade over SD card -- Nevo's memory system and QMD search index involve frequent small reads and writes that would bottleneck on flash storage within weeks.
Step 1: Physical Setup
Connect the Hardware
- Remove the Nevo Pi from its packaging
- Connect the Ethernet cable from the Pi to your router or network switch
- Connect the USB-C power supply to the Pi
- Plug the power supply into a wall outlet
That is it. The Pi will boot automatically when power is applied. The boot process takes approximately 30 seconds.
WiFi Alternative
If Ethernet is not practical, you can switch to WiFi after initial setup. We recommend Ethernet for reliability -- an AI agent running 24/7 should not depend on WiFi stability. To switch, SSH in and run:
sudo nmcli device wifi connect "YourNetworkName" password "YourPassword"
Verify with ip addr show wlan0, then disconnect the Ethernet cable.
Finding the Pi on Your Network
The Nevo Pi broadcasts its hostname as nevo-pi.local using mDNS. On most home and office networks, you can reach it immediately:
ping nevo-pi.local
If mDNS is not available on your network (common in some enterprise environments), check your router's connected devices list for a new device named "nevo-pi" and note its IP address.
Step 2: SSH Access and Initial Configuration
Nevo Pi is a headless device -- no monitor, no keyboard, no mouse. You interact with it over SSH from any computer on your network.
Connect via SSH
Open a terminal on your computer (Terminal.app on macOS, Windows Terminal on Windows, or any Linux terminal) and connect:
ssh nevo@nevo-pi.local
The default password is nevo-setup. You will be required to change it on first login.
Change Your Password
On first SSH login, the system will prompt:
Welcome to Nevo Pi. For security, please set a new password.
Current password: nevo-setup
New password: ******
Confirm password: ****
Password updated successfully.
Choose a strong password. This is the only authentication layer between your network and your AI agent.
Set Your Timezone
The setup wizard will ask for your timezone:
Select your timezone:
> America/New_York
America/Chicago
America/Denver
America/Los_Angeles
Europe/London
(type to search...)
Accurate timezone configuration matters for Nevo's scheduling, memory timestamps, and session management.
Step 3: Configure Your API Key
Like the desktop Nevo App, Nevo Pi uses Anthropic's Claude models for reasoning. You need an API key.
Get Your Anthropic API Key
If you do not already have one:
- Go to console.anthropic.com
- Create an account or sign in
- Navigate to API Keys
- Click Create Key and name it "Nevo Pi"
- Copy the key
Enter the Key
The first-login wizard will prompt for your API key:
Enter your Anthropic API key: sk-ant-***********************
Verifying... Connected to Anthropic API successfully.
API key saved to ~/.nevo/credentials/anthropic.env
You can also set the key later or update it:
nevo config set anthropic_api_key sk-ant-your-key-here
API Cost Expectations
Nevo Pi uses the same model routing as the desktop app. Simple tasks go to cheaper models, complex reasoning goes to Opus. Typical costs for moderate usage are $2-10 per day. Running Nevo Pi 24/7 does not continuously consume API credits -- it only makes API calls when you send it a task or a scheduled job triggers.
The Pi itself costs approximately $0.02/day in electricity. The total cost of ownership is your API spend plus two cents.
Step 4: Access the Web Dashboard
Nevo Pi includes a web-based dashboard you can access from any device on your local network -- your laptop, your phone, your tablet. No app installation required.
Open the Dashboard
In any web browser, navigate to:
http://nevo-pi.local:5174
Or use the IP address directly if mDNS is not available:
http://192.168.x.x:5174
Dashboard Overview
The dashboard gives you a real-time view of your Nevo Pi:
- Home -- System status, active tasks, recent activity feed, memory stats
- Agents -- The full roster of 14 specialized agents with status and performance metrics
- Tasks -- Send new tasks, view active and completed tasks, inspect quality pipeline results
- Memory -- Browse accumulated knowledge: facts, session narratives, rules, and skills
- Settings -- API key management, notification preferences, connector configuration
- Health -- CPU, memory, disk space, API connectivity, agent status, pipeline throughput
The dashboard is responsive. You can send Nevo a task from your phone, check progress from a tablet, or do deep inspection from your laptop.
Step 5: Connect Communication Channels
For day-to-day use, most people prefer messaging over the dashboard -- sending tasks from Telegram, Discord, or Slack the same way you would message a colleague.
To set up Telegram (the most popular connector), go to Settings > Connectors > Telegram in the dashboard, follow the guided setup, and send a test message from your phone. Once configured, you can message Nevo from anywhere. The message arrives at your Pi, gets processed locally, hits the Claude API for reasoning, and the response comes back to your phone. No static IP or port forwarding needed.
Discord, Slack, WhatsApp, and 15+ additional platforms follow the same pattern. You can run multiple connectors simultaneously. For more on why running AI privately matters, see Private AI Agents.
Step 6: Running 24/7
Nevo Pi is designed to run continuously. The system is configured out of the box for always-on operation, but understanding what is happening under the hood helps you maintain it.
The systemd Service
Nevo runs as a systemd service that starts automatically on boot and restarts automatically if it crashes:
# Check service status
sudo systemctl status nevo
# View recent logs
sudo journalctl -u nevo --since "1 hour ago"
# Restart the service (if needed)
sudo systemctl restart nevo
The service is configured with automatic restart on failure, a 10-second restart delay, and a watchdog timer that detects hangs. If Nevo becomes unresponsive for any reason, systemd will restart it within 15 seconds.
Health Monitoring
Nevo Pi runs its own health checks every 5 minutes. These verify:
- API connectivity to Anthropic
- Memory pipeline operation
- QMD search index integrity
- Disk space (alerts at 90% usage)
- Agent availability
- Service uptime
If a health check fails, Nevo will attempt automatic remediation. If remediation fails, you will receive a notification through your configured communication channel (Telegram, Discord, etc.).
You can also run a manual health check at any time:
nevo health --full
Or check from the dashboard under the Health tab.
Automatic Updates
Nevo Pi checks for updates daily. By default, security patches and minor updates apply automatically; major updates require confirmation through the dashboard. Updates never interrupt active tasks.
Power and Network Resilience
If power is lost and restored, the Pi boots and Nevo starts within 30 seconds. Interrupted tasks are flagged as incomplete and can be resumed. If network drops, Nevo pauses API-dependent work and resumes when connectivity returns. Local operations continue uninterrupted.
Step 7: Your First Task
With Nevo Pi running, verify the full pipeline. From the dashboard's Tasks tab or from Telegram, try:
Create a simple Python script that prints "Hello from Nevo Pi" and save it to ~/test-hello.py. Then run it and show me the output.
This tests file system access, code generation, script execution, and response delivery. If you see "Hello from Nevo Pi" in the output, every critical subsystem is working. For a deeper check, ask Nevo to list its active agents and capabilities -- this confirms the full agent roster is loaded.
Advanced Configuration
SSH key authentication -- For security, switch from password to key-based access. Run ssh-keygen -t ed25519 on your computer, then ssh-copy-id nevo@nevo-pi.local to deploy the key. Disable password auth in /etc/ssh/sshd_config for a locked-down setup.
Custom hostname -- If you run multiple Pis or prefer a different name: sudo hostnamectl set-hostname my-nevo-agent. The dashboard moves to http://my-nevo-agent.local:5174.
Remote access** -- Nevo Pi is designed for local network use. Do not expose it directly to the internet. For remote access, use a VPN -- Tailscale is recommended for simplicity.
Troubleshooting
Cannot Find the Pi on the Network
- Verify the Ethernet cable is connected and the link light on the Pi is active (solid green)
- Try pinging by hostname:
ping nevo-pi.local - If mDNS is not working, check your router's DHCP client list for a device named "nevo-pi"
- As a last resort, connect a monitor and keyboard to the Pi to check the IP address with
ip addr
SSH Connection Refused
- Ensure the Pi has finished booting (wait 60 seconds after applying power)
- Verify you are using the correct username:
nevo(notpiorroot) - Check that your computer is on the same network as the Pi
Dashboard Not Loading
- Confirm the Nevo service is running:
ssh nevo@nevo-pi.local "sudo systemctl status nevo" - Check port 5174 is not blocked by your network:
curl http://nevo-pi.local:5174 - Try accessing by IP address instead of hostname
- Restart the service:
ssh nevo@nevo-pi.local "sudo systemctl restart nevo"
Slow Performance
- Nevo Pi is an orchestration device, not a GPU cluster. Response times depend on the Anthropic API latency, typically 2-15 seconds per agent step
- If local operations (file browsing, memory search) are slow, check disk usage:
df -h. The NVMe should have ample free space - Monitor CPU with
htopvia SSH. Sustained 100% CPU suggests a runaway process -- restart the service
Overheating
- The aluminum case provides passive cooling sufficient for normal operation
- In very hot environments (ambient temperature above 35C / 95F), the Pi may throttle. Check with:
vcgencmd measure_temp - If temperatures consistently exceed 80C, improve airflow around the case or add a small USB fan
Frequently Asked Questions
Does Nevo Pi replace my computer?
No. Nevo Pi is a dedicated AI agent device that complements your computer. It handles AI workloads so your main machine stays fast and available for your own work. You interact with Nevo Pi through its web dashboard or messaging apps -- you never need to "use" the Pi directly.
Can I build my own Nevo Pi instead of buying the kit?
Nevo Pi is available as a pre-configured appliance for $200 because the value is in the software and configuration, not the hardware. Building your own from components is possible but not officially supported. The Nevo App ($100) is designed for macOS with Apple Silicon. A Raspberry Pi self-build guide may be available in the future for advanced users who want to source their own hardware.
How much electricity does Nevo Pi use?
The Raspberry Pi 5 draws approximately 5-12 watts depending on load. At average US electricity rates, that is roughly $0.50-1.00 per month. This makes it one of the most cost-effective ways to run a 24/7 AI agent.
What happens if my internet goes down?
Nevo Pi pauses any task that requires the Anthropic API and resumes automatically when connectivity returns. Local operations -- memory consolidation, file management, QMD indexing -- continue without internet. No data is lost during an outage.
Can I run multiple Nevo Pi devices?
Yes. Each Nevo Pi operates independently with its own memory, rules, and skill set. You can dedicate different Pis to different projects or workloads. Give each one a unique hostname during setup to distinguish them on your network. Multi-device orchestration (where one Pi delegates to another) is on the roadmap but not yet available.
Next Steps
- What Are AI Agents? -- Foundational guide to understanding AI agents
- Private AI Agents -- Why running AI on your own hardware matters
- Autonomous AI Agents -- How agents that work without supervision are built
- nevo.systems -- Order your Nevo Pi and join the community