Securely reach and monitor every machine you run β from one binary. No open ports, and nothing sensitive ever leaves your box.
You run things on machines you don't always sit in front of β Home Assistant on a Pi, a NAS, Pi-hole, Grafana, your own apps on an N100 or a VPS, a client's server. Beacon does two things first: it gives you secure remote access to those machines and the services on them β with no open ports β and it monitors their health and alerts you when something breaks. It's workload-agnostic: it reaches and watches whatever is on the box, not just things it deployed. (It can run your deploys too, but access and monitoring are the core.)
And it's built on a strict local-first, zero-trust model β your code, secrets, and data never leave the box, and no cloud ever holds root on your machines.
⬑ beacon v0.3.2 β running pid 1847 uptime 14d 3h
DEVICE pi-homelab 192.168.1.42 arm64 Debian 12
SYSTEM cpu 12% ββββββββββββββββ mem 67% ββββββββββββββββ disk 41% ββββββββββββββββ
load 0.42 0.38 0.35 temp 48Β°C
PROJECTS 3 healthy 1 warning 0 down
β portfolio-site v2.1.0 deployed 2h ago 3/3 checks passing
β home-assistant v2024.3 deployed 5d ago 2/2 checks passing
β nextcloud v28.0.1 deployed 3d ago 2/3 checks passing
ββ β HTTP https://cloud.local/status timeout 5.2s > 3s threshold
That's beacon status. There's also a browser dashboard at http://localhost:9100 β self-contained, no CDN, works offline.
Two things first β secure remote access and monitoring β on a zero-trust foundation. Everything else is a bonus.
This is the part that matters before any feature:
- No open ports. Tunnels connect outbound from your device β nothing is exposed to the internet, and it works behind CGNAT.
- No control plane with root. The optional cloud can never log in to your machine on its own; it only relays commands you authorized, and the agent runs them.
- Nothing sensitive leaves the box. Your source code, deploy scripts, tokens, and secrets stay local. The cloud only ever sees the metrics and log lines you explicitly choose to send.
- Works offline. The agent and its dashboard run fully local, with no account.
- Reach any service, no open ports β securely reach Home Assistant, Grafana, Jellyfin, a NAS admin page, Pi-hole, or any local HTTP service from anywhere. No dynamic DNS, no Nabu Casa, no Cloudflare account or managed domain. Authenticated with short-lived tokens; only you can reach your services.
- Remote terminal β open a shell on your device from the browser. No SSH port, no VPN. The cloud relays a PTY session between your browser and the agent.
- Device-verified gate (recommended) β enroll a passkey (preferred) or set a passphrase (fallback) so a second factor is required before any remote terminal or tunnel session can open. It's verified on the device β BeaconInfra only relays. The passkey's private key never enters the page, and you can layer on an authenticator-app (TOTP) code so the cloud can't silently open a session. See Securing remote access.
- Health & metrics β health checks (HTTP, port, command), CPU/memory/disk/temperature, per-project status, Prometheus metrics.
- Alerts β via webhook (Slack, Discord) or SMTP, with quiet hours and offline-device detection.
- Log forwarding β tail log files, Docker container logs, or
journalctland forward them to the dashboard, filtered so you only ship what matters.
- Deploy automation (bring your own script) β Beacon doesn't replace your stack or turn your box into a PaaS. Point it at a Git repo or Docker registry; it watches for new tags, injects your secrets, and runs your deploy script. Push a tag, walk away.
- WireGuard VPN β turn any Beacon device into a peer-to-peer WireGuard exit node and route traffic through your home network from a laptop.
BeaconInfra is the optional cloud that adds the multi-device dashboard, tunnels, and remote terminal. When an API key is configured, beacon start sends periodic heartbeats with device metrics and project health; the cloud uses these to power the dashboard, detect offline devices, and relay commands back to the agent. Everything except tunnels and remote terminal works without an account.
Beacon isn't a PaaS and doesn't replace how you deploy. It's the operations layer across all your machines β the box running your apps and the Home Assistant box, the NAS, the Pis, the client server, the VPS you never containerized β so you reach, watch, and fix every one of them from a single place, without handing any cloud root over your hardware.
If you run Home Assistant OS, the primary path is the Beacon Home Assistant add-on:
install the add-on, paste your BeaconInfra API key, enable tunnel_home_assistant,
and start it. The add-on points the tunnel at Home Assistant Core on
http://homeassistant:8123.
If you run Beacon on a normal Linux host or in Docker next to Home Assistant, use the CLI tunnel flow. Three commands, no port-forwarding, no Nabu Casa.
# 1. Log in to BeaconInfra (free account)
beacon cloud login --api-key usr_xxxxxxxx
# 2. Expose Home Assistant
beacon tunnel add homeassistant --port 8123
# Or, for Docker Compose / HA OS style service DNS:
# beacon tunnel add homeassistant --port 8123 --host homeassistant
# 3. Start Beacon
beacon startYour Home Assistant is now accessible from the BeaconInfra dashboard β on your phone, from a hotel, wherever. The tunnel connects outbound from your device (no inbound ports needed), reconnects automatically, and works behind CGNAT.
The same tunnel works for Grafana, Jellyfin, Pi-hole, Nextcloud, your NAS admin page, a staging server β anything that speaks HTTP on your LAN.
Home Assistant setup: HA blocks proxied requests by default. Add this to your configuration.yaml (the file inside your HA config volume):
http:
use_x_forwarded_for: true
trusted_proxies:
- 172.30.0.0/16
- 172.16.0.0/12
- 127.0.0.1Then restart Home Assistant Core. Without this, the tunnel connects but HA can return 400 Bad Request or show "can't connect to Home Assistant" because it rejects the forwarded proxy headers.
Remote terminal and tunnel sessions are reached through the BeaconInfra relay. By default the relay is gated by short-lived, authenticated tokens. If you want a second factor that the cloud cannot forge or replay, turn on the remote-access gate. Once configured, no remote terminal or tunnel session can open until you unlock it, and the unlock is verified on the device β the cloud only relays.
A passkey (WebAuthn) is the preferred factor; a passphrase is the fallback. On top of either, you can add an authenticator-app code (TOTP) as an out-of-band confirmation. Enroll whichever you like.
# Preferred: enroll a passkey (Touch ID / Windows Hello / security key / phone).
# Prints a one-time enrollment code to paste into the dashboard.
beacon remote-access add-passkey
beacon remote-access list-passkeys
beacon remote-access remove-passkey <id-or-label>
# Fallback: set (or change) a passphrase β prompted twice, no echo
beacon remote-access set-passphrase
# Optional out-of-band factor: scan the QR into an authenticator app.
# Every unlock then also asks for the current 6-digit code.
beacon remote-access setup-oob
beacon remote-access remove-oob
# Check what's configured
beacon remote-access status
# Remove everything (local recovery; disables the gate)
beacon remote-access clearRestart beacon after changing factors so a running agent picks up the gate.
In both factors, your passphrase never travels over the network. The browser keeps it local, derives a key from it, and sends only a one-time computed proof β the agent verifies that proof on the device. The cloud never sees the passphrase or anything reusable.
A passkey is simply safer still: there's no typed secret at all. Its private key lives in your authenticator hardware (Touch ID, Windows Hello, a security key, your phone) and is used to sign a one-time challenge; the agent stores only the public key and verifies the signature. Nothing the browser handles could be captured and replayed later, which is why passkeys are the preferred factor.
- Run
beacon remote-access add-passkeyon the device. It prints a one-time enrollment code (also visible in the Home Assistant add-on log), valid for a few minutes and usable once. - In the dashboard, choose Add passkey, enter the code, and complete your browser's passkey prompt (biometric / PIN).
- The browser creates the passkey bound to the dashboard's domain (
rpId) and sends the public key to the device. The agent verifies the registration and stores only the public key, credential id,rpId/origin, and a signature counter in~/.beacon/remote-access.json(mode0600).
Rooting enrollment in a device-local code (or a valid passphrase unlock) means a compromised cloud cannot enroll its own passkey β it never has the code.
To add the optional authenticator-app factor, run beacon remote-access setup-oob.
It generates a TOTP secret and prints an otpauth:// QR code (and the plain
secret) right in the terminal β or the Home Assistant add-on log. Scan it with any
authenticator app (Google Authenticator, 1Password, Authy). The secret is created
on the device and stored locally; it never reaches the cloud.
Opening a terminal or tunnel is verified on the device:
- A WebAuthn assertion. The agent issues a single-use, short-lived nonce
as the challenge. Your authenticator signs it (user verification β biometric or
PIN β is required). The agent verifies the signature against the stored
public key, checks the
rpId/origin it pinned at enrollment, the nonce, the user-present/user-verified flags, and a monotonic signature counter (replay defense). With the passphrase fallback, the browser instead returns a proof =HMAC-SHA256(Argon2id-key, nonce β action β session_id), compared in constant time. Either way the cloud sees no reusable secret. - The authenticator-app code, when enrolled. If you ran
setup-oob, the unlock also asks for the current 6-digit TOTP code from your app. The agent verifies it against the locally-stored secret; the code is single-use within its window. This is generated on your phone and checked on the device β the cloud is never in the loop.
A successful unlock is in-memory, single-use, session-bound, and TTL-limited, and is cleared on restart (fail-closed). Repeated wrong attempts (assertion or code) trigger rate-limiting / backoff. Tunnels do not auto-start while the gate is configured β each comes up only on an unlocked connect, so a restart (which clears grants) requires unlocking again before any local service is reachable.
- No reusable secret crosses the network. Only a one-time computed proof (or a passkey signature) is sent, and the device verifies it β there's nothing to capture and replay later.
- The cloud cannot silently open a session. Unlocking always takes a deliberate action on your own hardware β a biometric touch on your authenticator, and (when enrolled) the current code from your authenticator app, which the cloud never sees.
With no factor configured, behavior is unchanged (the gate is off, and tunnels auto-start as before).
curl -fsSL https://raw.githubusercontent.com/Bajusz15/beacon/main/scripts/install.sh | bashOne static binary, no runtime dependencies. Builds for Linux (AMD64, ARM64, ARMv7) and macOS.
beacon init --name my-piWrites ~/.beacon/config.yaml with your device name. No network calls, no account needed.
beacon startDashboard at http://localhost:9100. System metrics, project health, Prometheus endpoint β all running locally.
beacon cloud login --api-key usr_xxxxxxxx
beacon start # restart to enable heartbeats + tunnelsThe first heartbeat registers your device automatically. To disconnect: beacon cloud logout. Beacon makes zero outbound calls without an API key.
Beacon manages your apps end-to-end: clone from Git or pull from Docker, run your deploy command, poll for updates, health check, and tail logs. Each project runs as its own isolated process β one crash doesn't affect others.
beacon bootstrap myappThe wizard asks for deployment type (Git or Docker), repo URL, tokens, and deploy command. It creates a systemd service and kicks off the first deployment.
beacon bootstrap myapp -f bootstrap.ymlGit:
deployment_type: "git"
repo_url: "https://github.com/you/myapp.git"
git_token: "ghp_xxxxxxxxxxxx"
local_path: "$HOME/beacon/myapp"
deploy_command: "./scripts/deploy.sh"
poll_interval: "60s"Docker:
deployment_type: "docker"
local_path: "$HOME/beacon/myapp"
poll_interval: "60s"
docker_images:
- image: "ghcr.io/you/web-app"
token: "ghp_xxxxxxxxxxxx"
deploy_command: "docker compose up -d"
docker_compose_files:
- "docker-compose.yml"Beacon talks to the registry API, detects the newest tag, pulls it, and runs your command. Supports Docker Hub, GHCR, and any Registry v2-compatible registry. Multiple images in one project are tracked independently β only the changed image redeploys.
Beacon can store encrypted deploy-time secrets per project and environment:
beacon secrets set API_TOKEN --project myapp --env prod
beacon secrets list --project myapp --env prod
beacon secrets list --reveal --project myapp --env prod
beacon secrets export --reveal --project myapp --env prodSecrets live under ~/.beacon/secrets/<project>/<env>.enc and are encrypted with a local AES-256-GCM machine key at ~/.beacon/secrets/key. The key is created automatically with 0600 permissions.
During deploys, Beacon starts from the process environment, loads the existing BEACON_SECURE_ENV_PATH .env file when configured, then overlays Beacon secrets. That means Beacon secrets override matching .env values. Set BEACON_PROJECT_ENV=prod in the project env file to select an environment; if it is not set, Beacon uses default.
Values are not printed unless you pass --reveal. Use list for key names only, list --reveal to inspect all values, and export --reveal --format env|json for scripts.
Security note: the machine key is stored locally next to the encrypted secret files. This protects against accidental commits, backup leakage, and casual inspection; it is not meant to protect against someone who can already read your Beacon home directory.
# ~/.beacon/config/projects/myapp/monitor.yml
checks:
- name: "http_200"
type: http
url: "http://localhost:8080/health"
interval: 30s
- name: "process_running"
type: process
name: "myapp"# ~/.beacon/config/projects/myapp/alerts.yml
channels:
- name: slack
type: webhook
url: "$WEBHOOK_URL"
routing:
- severity: critical
channels: [slack]
- severity: warning
channels: [slack]
quiet_hours:
start: "23:00"
end: "07:00"
timezone: "Europe/Budapest"Test it: beacon alerts test --project myapp --severity critical
Open a shell on your device from the BeaconInfra dashboard β no SSH, no VPN, no open ports.
The agent picks up a terminal_open command via heartbeat, dials back to the cloud over WebSocket, and spawns a local PTY shell. Browser β Cloud β Agent, end-to-end. Sessions auto-expire after 15 minutes or 5 minutes idle.
Security: one-time tokens per session (SHA-256 hashed, server stores only the hash), shell restricted to a known allow-list (bash, zsh, sh, fish, etc.), runs as the Beacon agent's OS user.
Turn any Beacon device into a peer-to-peer WireGuard exit node. Your traffic flows directly between devices β BeaconInfra only coordinates the key exchange and endpoint discovery.
# Home device (exit node β needs one port-forwarded UDP port)
beacon vpn enable --listen-port 51820
# Laptop (anywhere)
beacon vpn use my-home-piUse case: you're on airport WiFi and want to route through your home connection. No subscription, no third-party relay, no trust required. WireGuard is cryptographically silent β port scanners can't even tell it's listening.
Tear down: beacon vpn disable.
Beacon does a lot in one binary. The tables below are a quick tour β if something looks useful, the sections above cover the full setup.
Everything here works without an internet connection and without signing up for anything.
| You want to⦠| How |
|---|---|
| Deploy a Git repo automatically when you push a new tag | beacon bootstrap myapp β point it at your repo, give it your deploy script. Beacon polls for new tags and runs the script. |
| Deploy Docker images automatically from Docker Hub, GHCR, or any private registry | Use deployment_type: docker in your bootstrap config. Beacon watches for new tags and runs your docker compose up -d (or anything else). |
| Deploy a whole stack where each image moves independently | List multiple images under docker_images: in your bootstrap. Only the image that changed redeploys. |
| Check that an HTTP endpoint is up | Add an HTTP check to your project's monitor.yml β set a URL, an interval, and a timeout. |
| Check that a port is open (databases, SSH, custom services) | Add a type: port check with a host and port. |
| Check anything a shell command can check | Add a type: command check. The exit code tells Beacon if it's up. |
| See everything at a glance from the terminal | beacon status β colored summary of every project. Add --watch for a live view. |
| See everything in a browser | Open http://<your-device>:9100. Self-contained dashboard that auto-refreshes. |
| Pull metrics into Grafana / Prometheus | Scrape http://<your-device>:9100/metrics. |
| See CPU, memory, disk, load, temperature | Enabled by default. Shows up in beacon status and the dashboard. |
| Get a Slack / Discord / webhook message when something goes down | Create alerts.yml next to your monitor.yml. Route by severity to any webhook. |
| Get an email when something goes down | Same alerts.yml, add an email channel with your SMTP details. |
| Silence alerts at night | Add quiet_hours: to your alert routing with a start/end time and timezone. |
| Test your alert setup without waiting for an outage | beacon alerts test --project myapp --severity critical |
Forward logs from a file, a Docker container, or journalctl |
Add a log_sources: block to your monitor.yml. Filter with include/exclude patterns so you only ship what you care about. |
| Keep your tokens out of config files | beacon keys add β encrypted local token store for Git, Docker, webhooks. |
Keep deploy secrets out of .env files |
beacon secrets set API_TOKEN --project myapp --env prod. Beacon injects them after .env values during deploy. |
| Access Home Assistant, Grafana, or any local service remotely (with a BeaconInfra account β authenticated, no port-forwarding needed) | beacon tunnel add homeassistant --port 8123 |
| Run several tunnels at once | beacon tunnel list / beacon tunnel enable / beacon tunnel disable |
| Route traffic through your home network from your laptop | beacon vpn enable on the exit node, beacon vpn use <device> on the client. Peer-to-peer WireGuard. |
| Query Beacon from Cursor or Claude Desktop | beacon mcp serve β see docs/MCP.md |
| Monitor a Kubernetes cluster | beacon source add with your kubeconfig. |
| Manage your project list | beacon projects list, beacon projects add, beacon projects status myapp |
A free BeaconInfra account adds a hosted dashboard and remote access on top of everything above. Your device keeps running locally β the cloud just gives you somewhere to see it all from a browser, including from your phone.
Turn it on with beacon cloud login --api-key usr_β¦. Turn it off any time with beacon cloud logout.
| You want to⦠| What you get |
|---|---|
| See all your devices in one place | One dashboard showing every machine running Beacon β your Pi, your NAS, your VPS, your homelab server β with current health, uptime, and system metrics. |
| Open Home Assistant from your phone, anywhere | Set up the homeassistant tunnel once. From then on, open the BeaconInfra dashboard on your phone and click through to your HA UI. No VPN, no port-forwarding, no dynamic DNS. |
| Reach any other local service remotely | The same tunnel mechanism works for Grafana, Jellyfin, Pi-hole, your router's admin page, a staging VM β anything that speaks HTTP on your LAN. |
| View logs from anywhere | The log lines you configured to forward show up in the dashboard, filterable by device and project. Useful when something breaks and you don't want to SSH in. |
| Watch your metrics remotely | CPU, memory, disk, load, and temperature for every device β without being on the LAN. |
| See all your project health in one list | Every project, every check, across every device. Sorted so the problems come first. |
| Trigger a deploy from the browser | Click "deploy" in the dashboard and Beacon runs your existing deploy script on the device. Your secrets never leave home β the cloud just sends the signal. |
| Know when a device goes offline | If a device stops sending heartbeats, you get notified β even if its last check said everything was fine. |
| Open a remote terminal session | Click "Open terminal" on a device in the dashboard. The cloud relays a shell session (PTY) between your browser and the Beacon agent β no SSH port, no VPN needed. Sessions auto-expire after 15 minutes. Enroll a remote-access passkey or passphrase to require a device-verified second factor before any session opens. |
| Route traffic through your home network | beacon vpn enable on your home device + beacon vpn use my-pi on your laptop. WireGuard peer exchange happens via BeaconInfra; the actual traffic is peer-to-peer. For client-only machines, use the lightweight beacon-vpn binary. |
Even with BeaconInfra enabled, some things stay on your device and never touch the cloud:
- Your source code and deploy scripts β the cloud only sends a "deploy now" signal; your device runs the script.
- Your tokens (Git, Docker, webhooks) β encrypted locally by
beacon keys. - Your application secrets (database passwords, API keys loaded via
BEACON_SECURE_ENV_PATHorbeacon secrets) β Beacon hands them to your app at deploy time and nothing else. - Raw log files β only the lines you explicitly configured as
log_sourcesare forwarded. Everything else stays on disk. - The local dashboard at port 9100 β it keeps working offline, BeaconInfra account or not.
If you change your mind, beacon cloud logout stops all outbound reporting on the next heartbeat. There's nothing to delete from a control panel because there's no persistent account state beyond what you chose to send.
beacon start runs one orchestrator process per device (the "master"). It collects system metrics, serves the local dashboard, sends heartbeats, and supervises everything else. It's stateless per project β it doesn't know about Docker or systemd.
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β BeaconInfra Cloud (optional) β
β heartbeats, commands, tunnel proxy, terminal relay β
ββββββββββββ¬ββββββββββββββββββββββββββββ¬ββββββββββββββββ
β HTTPS β WebSocket
ββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
β beacon start β
β β
β One per device. System metrics, local dashboard, β
β heartbeats, project supervision, tunnel + VPN mgmt. β
ββββ¬βββββββββββββββ¬βββββββββββββββ¬βββββββββββ¬βββββββββββ
β IPC β IPC β goroutine β WireGuard
ββββ΄ββββββββββββ ββ΄βββββββββββββ ββ΄ββββββββββ ββ΄ββββββββββ
β project agentβ βproject agentβ β tunnels β β VPN β
β myapp β β blog β β HA :8123 β β beacon0 β
β health checksβ β health checkβ β NC :8080 β β 51820/UDPβ
β log tailing β β log tailing β β (WS proxy)β β β
ββββββββββββββββ βββββββββββββββ βββββββββββββ ββββββββββββ
Projects are isolated: one crash doesn't affect others. The master auto-restarts failed projects with exponential backoff. Tunnels run as lightweight goroutines inside the master, connecting outbound to the cloud via WebSocket so local services are accessible without opening ports.
| Command | Purpose |
|---|---|
beacon start |
Start Beacon (dashboard, projects, tunnels, heartbeats) |
beacon status |
Terminal health view (--json, --watch, --no-color) |
beacon init |
Write local config (--name, --metrics-port; no network) |
beacon cloud login / logout |
Enable/disable cloud |
beacon bootstrap <name> |
Set up a project (interactive or -f config.yml) |
beacon deploy |
Git/Docker tag polling loop |
beacon secrets set|get|list|export|remove |
Local encrypted deploy secrets |
beacon tunnel add|list|enable|disable |
Reverse tunnels for remote access |
beacon remote-access add-passkey|list-passkeys|remove-passkey |
Enroll/manage passkeys (preferred) that gate remote terminal/tunnel sessions |
beacon remote-access setup-oob|remove-oob |
Enroll/remove an authenticator-app (TOTP) out-of-band factor |
beacon remote-access set-passphrase|clear-passphrase|status|clear |
Device-verified passphrase (fallback), gate status, and reset |
beacon vpn enable|use|disable|status |
WireGuard VPN |
beacon projects list|add|remove|status |
Project management |
beacon alerts init|test|status |
Alert routing |
beacon keys list|add|rotate|delete |
Encrypted token store |
beacon mcp serve |
MCP server for Cursor / Claude Desktop |
beacon config show |
Show resolved paths and identity |
beacon update |
Self-update to latest release |
beacon bootstrap installs systemd services automatically. For manual setup:
cat > ~/.config/systemd/user/beacon.service << 'EOF'
[Unit]
Description=Beacon Agent
After=network-online.target
Wants=network-online.target
[Service]
Type=simple
ExecStart=/usr/local/bin/beacon start
Restart=on-failure
RestartSec=30
[Install]
WantedBy=default.target
EOF
systemctl --user daemon-reload
systemctl --user enable --now beacon.service- docs/MASTER_AGENT.md β agent architecture and heartbeats
- docs/VPN.md β WireGuard VPN setup and security model
- docs/LOG_FORWARDING.md β log forwarding
- docs/KEY_MANAGEMENT.md β encrypted key store
- docs/MCP.md β MCP server for editors
- examples/ β bootstrap, monitor, alert configs
β Buy me a coffee β if Beacon saves you time.
Apache 2.0 β see LICENSE