Add each machine to fleet.yaml. You can monitor over SSH, or install the agent on the machine for outbound-only connectivity.
nodes:
- name: kids-pc
host: 192.168.1.131
ssh: monitoring@192.168.1.131
os: windows
Option A: Install the agent (recommended)
Install the lightweight Leassh agent on each device. Enter a pairing code and it connects automatically. Works through firewalls and NAT.
Windows (PowerShell as Admin)powershell -ExecutionPolicy Bypass -Command "& { $env:LEASSH_TOKEN='YOUR_TOKEN'; iwr -useb https://leassh.com/install-windows.ps1 | iex }"
Replace YOUR_CODE with the pairing code from your dashboard. For detailed platform-specific instructions, see the Install page.
Option B: SSH Probing (no install on targets)
Add machines to fleet.yaml with their SSH details. Leassh probes them directly. Nothing installed on the monitored machines. Requires SSH key access. See Self-Hosting for details.
Step 3: Add your license
Paid subscription licenses are online LSH-… keys validated through leassh.com with a 72-hour local cache. Offline LSH1.… Ed25519 tokens remain available for paid installations that need them. Free is an online 2-computer license issued at registration and works through the same path. See Plans and licensing.
Step 4: See it everywhere
The Home Assistant integration turns every node into a device with presence, load, and safe action buttons; the OpenClaw plugin exposes the same fleet to your agent; anything with HTTP speaks the REST+WS API below.
Plans and licensing
One catalog, four plans. Prices and limits are the same everywhere on the site and in the server entitlement table.
Plan
Computers
Annual / mo
Monthly
Trial
Free
2
$0
$0
—
Family
10
$9.99
$14.99
14 days
Pro
25
$19.99
$29.99
14 days
Unlimited
Unlimited
Contact
—
Free includes the local dashboard, Home Assistant integration, sensors, binary sensors, presence, curated buttons, notify, Assist, basic alerts and one ha API token. Family adds AI screen analysis, weekly reports, screen time, limits/enforcement, content alerts, and Telegram/email channels. Pro adds the fleet dashboard, rules, webhooks, MQTT, SSH actuation, the full REST API and OpenClaw, for 25 computers. Unlimited adds unlimited nodes, priority support and multi-site onboarding.
API tokens are declared under api.tokens. Free accepts exactly one token, and the documented primary scope is ha. Using Home Assistant and OpenClaw at the same time requires Family or Pro. The former entry tiers are retired; their IDs are accepted as legacy IDs mapped to Family for 90 days.
Connection Modes
Leassh supports two ways to connect machines. Both provide the same monitoring data.
Agent
SSH
Install on target
Yes (tiny service)
No
Setup difficulty
Enter pairing code
Configure SSH keys
Works through firewalls
Yes (outbound only)
Needs inbound SSH access
Screenshot capability
Yes (with macOS permission)
Yes (with macOS permission)
Remote command execution
Paid only (tunnel mode)
Yes
Best for
Desktops, laptops, remote machines
Servers, NAS, VMs
Deploy Agents via SSH
If you already have SSH access to a machine, you can deploy the agent remotely without physically touching the machine:
macOS Note: After deploying the agent on a Mac, the Screen Recording permission must be approved once from the machine's physical keyboard. Open System Settings → Privacy & Security → Screen Recording and enable “Leassh Capture”. This is a one-time step required by Apple — no software can bypass it. All other monitoring features work immediately without this step.
Self-Hosting
Run everything on your own hardware. Zero data leaves your network.
Option A: Binary (recommended)
Download the server binary from GitHub Releases and run it directly:
Linux x64curl -L https://github.com/cgoberg/releases/releases/latest/download/leassh-server-linux-x64.tar.gz | tar xz
./leassh fleet.yaml
Option B: Docker
The container image is hosted on GitHub Container Registry. Authenticate first:
The openclaw-fleet plugin connects your Leassh fleet to any OpenClaw-compatible AI agent. Ask natural language questions about your infrastructure and get real-time answers.
Install the plugingit clone https://github.com/cgoberg/openclaw-fleet.git
cd openclaw-fleet
pip install -r requirements.txt
v0.2.0 adds server mode: pair the plugin with a Leassh server ({"mode": "server", "serverUrl": "…", "apiToken": "…"}) instead of spawning a local binary — same tools, one fleet authority. Standalone mode now mDNS-probes and warns before running a second authority.
Once installed, you can ask your AI agent things like:
One dialect, multiple transports: the same JSON-RPC vocabulary over stdio, REST, and WebSocket. Scoped tokens (forge · ha · openclaw · admin) are minted in fleet.yaml; the HA scope structurally cannot execute commands.
REST
GET /api/health # liveness + license banner (public)
GET /api/devices # bulk fleet state (ETag/304)
GET /api/devices/{id} # node detail
POST /api/devices/{id}/action # curated verbs: reboot_node, sign_out_session
POST /api/notify # {source, level, title, body, ack_url?}
GET /api/actions # the curated verb catalog
Events: node.online/offline/activity, trigger.*, notify, action.*. The server advertises _leassh._tcp via mDNS for discovery.
Home Assistant
The Home Assistant integration is included on every plan, including Free. Add cgoberg/leassh-integration as a custom repository in HACS, or use the HAOS add-on which installs and pairs everything automatically. Every fleet node becomes a device — presence, signed-in, CPU, RAM — with curated action buttons and urgent notifications as iOS critical alerts through your own HA.
Free contract: 2 nodes, the local dashboard, sensors, binary sensors, presence, curated buttons, notify, Assist, and exactly one ha API token. A third node is rejected with a clear NODE_LIMIT_REACHED error; existing nodes keep reporting if a downgrade leaves you over the limit. License errors surface as an HA repair (license_invalid, license_expiring, or node_limit_reached) rather than “cannot connect”.
The add-on accepts both online LSH-… keys and offline LSH1.… tokens. Its working directory is /config/leassh, so data/license_cache.json and screen-time data survive restarts.
FAQ
Common questions are answered on the main site FAQ section. Here are some additional technical questions:
Does Leassh require open inbound ports?
No. The agent communicates outbound to your Leassh server. No inbound ports need to be opened on monitored machines. For self-hosted setups, you need the server to be reachable from agent machines on your LAN.
Can I run Leassh without Ollama?
Yes. Without a vision model, Leassh still captures process lists and system metrics. Screenshot classification will be disabled, but time limits based on process names still work.
What happens if the agent loses connection?
The agent buffers data locally and syncs when the connection is restored. Screen time enforcement continues to work offline based on the last known rules.
Can I monitor machines outside my LAN?
Yes. With the agent installed, machines connect outbound through any firewall or NAT. With SSH probing, use SSH tunneling or a VPN to reach machines outside your LAN.
Do I need to install software on the computers I'm monitoring?
You have two options. The recommended approach is to install the tiny Leassh agent (under 10MB) — just enter a pairing code and it connects automatically. Alternatively, power users can use SSH probing which requires no software on target machines but needs SSH access configured. Both give you the same monitoring data.