Vigil watches your machines and, when you want it to, changes them. A small agent on each host checks in with a central server. The server collects metrics, raises alerts, keeps a hardware and software inventory, finds vulnerabilities, and sends signed tasks back to the agents to fix things: patch an app, restart a service, clear a disk, rebuild a machine.
It runs on Linux, Windows and macOS hosts, installs with docker compose up,
and has no limit on hosts, agents or retention.
Every screenshot below is a demo instance with made-up hosts and data.
| See | Live dashboards, per-host monitoring, alerts, a hardware and software inventory, a network map |
| Act | Tasks in YAML, playbooks with if/else branches, automations on events and schedules, wave-by-wave rollouts |
| Patch | App and update policies with maintenance windows, approvals and SLA tracking |
| Secure | Vulnerability management with one-click fixes, firewall management, signed tasks, two-factor approval |
| Operate | Docker and Podman, remote reprovisioning, public status pages, Active Directory import |
The dashboard is a grid of widgets you arrange, resize and name. Keep several (an overview, a wall display, one per client) and share them. The host cards show status, OS, tags, agent mode, and live CPU, memory, disk and network.
The Monitor page opens one host: gauges, load, and history from an hour to a week.
The Inventory is a sortable, filterable table of every host: manufacturer, model, CPU, RAM, BIOS, disks, MAC addresses, uptime, last logged-in user. Drag columns into order, export to CSV, and add your own columns from the output of a task. Tag hosts however you like and act on a whole tag at once.
The Apps page lists every piece of software your hosts report, by app, by host, or by the update it is missing. It reads dpkg, rpm, Flatpak, Snap, winget, Chocolatey, Scoop and the Windows registry, and flags what is outdated or unmanaged.
Twenty alert rules ship ready to use, with auto-resolve and host-offline detection. Alerts go out by webhook or email, can open tickets, and can be acknowledged or snoozed. Suggest Fix drafts a remediation task for you to review before it runs, using an AI model of your choice (bring your own key, any OpenAI-compatible endpoint or Anthropic) when one is needed.
A task is a short YAML file: a name, a risk level, and a list of actions
such as restart_service, update_package, clear_temp_files or
hunt_process. Deploy one to a host, a tag, or the whole fleet. Tasks can have
schedule windows, retries, success criteria, typed inputs asked for at deploy
time, and a collect: block that writes their output into the inventory.
Detection tasks hunt for files, packages, processes and ports. Browse and
submit to the community catalog.
A playbook runs tasks in order on a host, and it can decide what to do next. Add if / else branches that read an earlier step's status or the values its task reported, nest them up to three deep, let a step fail and carry on, and pass inputs down from the playbook to its steps. The page draws the flow as you build it.
This one finds out whether nginx is running. If it is, it updates it; if not, it installs it. Then it verifies, and if the verify fails it reloads and tries once more.
Branches nest, so a fallback can have its own decision inside it:
A playbook can run on its own when a matching host is approved (auto-enroll),
or be called from any task with type: playbook. Templates give you a head
start.
An automation runs a task or a playbook when something happens, or on a schedule. Triggers are events (an alert is sent, a host is approved) filtered by severity, alert rule, text in the alert, or tag, or a cron schedule. Run it on the host the event came from, on a tag, or on everything, either all at once or wave by wave with automatic halting if too many hosts fail.
A policy says what a set of hosts should have: each app present, latest, pinned to a version, or absent, plus which OS updates to install, how long to defer them, and when hosts may restart. Vigil compares each host against its policy and fixes drift only inside the maintenance window. Turn on approve each change and the changes wait for a person instead. A compliance view tracks the percentage of hosts patched within your SLA.
The By update view on the Apps page groups what hosts are missing by KB or package. Approve or decline one update and it overrides every policy at once.
Findings come from Trivy, Nessus and Greenbone, plus Vigil's own matcher against OSV, CISA KEV and EPSS that works offline. Findings are grouped by the fix that resolves them, so one Deploy fix patches every affected host at once.
Read a host's firewall (ufw, firewalld, or Windows Firewall) on demand, change the default policies, and add or remove rules. The page refuses the changes that would lock you out: blocking the agent's own outbound connection, closing SSH (or RDP on Windows), or defaulting incoming traffic to deny with no rule for it.
The Topology page groups your hosts by site and shows each one's address, open ports, firewall state, compose stacks, and which hosts talk to which.
Docker and Podman containers and compose stacks across your hosts: start, stop,
restart, read logs, update images, roll back, and edit a stack and its .env
in place.
Reprovision keeps a library of OS images (Ubuntu, Debian and RHEL-family), install profiles that answer everything the installer asks (disk, partitioning, network, admin user, packages, tags to apply when done), and rebuild jobs that reinstall a host unattended and bring it back enrolled. It destroys the target disk, so it asks first.
Share a public status page from the same data your monitoring already has. Choose which hosts appear, give them customer-facing names, and the page shows a 90-day uptime history for each. The first page is free; branding and per-site pages are part of Business.
- Active Directory import with automatic tags from OU paths.
- Sites, roles and an audit log (Business) to separate offices or clients, scope who can do what, and record who changed what.
- Notifications by webhook or email, and Jackil to turn an alert into a ticket.
- Agent modes so each host decides how much Vigil may do to it.
- Every task is signed with Ed25519. Agents pin the server key on first use and refuse tasks if it ever changes.
- Each agent runs in one of three modes, monitor, managed or full control, and the agent's own config is the final say. A compromised server cannot raise an agent's privileges. Monitor mode runs unprivileged.
- Risky steps and host approval can require a TOTP code, single-use with replay protection.
- No
shell=Truein the agent, and every task parameter is validated before it runs.
You need Docker with the Compose plugin.
git clone https://github.com/Susquehanna-Syntax/Vigil.git
cd Vigil
make setup # copies .env.example to .env
make gen-key # prints a signing key seed
make gen-secret # prints a Django secret keyOpen .env and paste in VIGIL_SIGNING_KEY_SEED, DJANGO_SECRET_KEY, and a
POSTGRES_PASSWORD. Then:
make up # starts Django, TimescaleDB, Redis, Celery
make superuser # create your admin loginOpen http://localhost:9200 and sign in.
Pasting the stack into Portainer, or any UI that round-trips through YAML? Read the note on
VIGIL_SIGNING_KEY_SEEDin the reference first. A mangled seed breaks every agent check-in.
-
In the dashboard click Add Host, pick an operating system and a mode.
-
Copy the one-line install command it shows and run it on the machine:
# Linux / macOS curl -fsSL https://vigil.example.com/agent/install.sh | sudo env VIGIL_TOKEN=<token> bash
# Windows, in an elevated PowerShell $env:VIGIL_TOKEN = "<token>"; irm https://vigil.example.com/agent/install.ps1 | iex
-
The host appears in Settings → Enrollment Queue. Approve it and metrics start flowing within a few seconds.
Start in monitor mode. You can move a host to managed or full control later with one config line and a re-run of the installer.
Tasks are plain YAML. This one reloads nginx and checks that it came back:
name: Reload nginx
description: Reload nginx and confirm it is active.
risk: low
actions:
- id: reload
type: reload_service
params:
service_name: nginx
- id: verify
type: check_service
params:
service_name: nginx
expect: activePaste it into Tasks, pick a host or a tag, and deploy.
For a quick look on a laptop, Vigil runs on SQLite with no other services:
cd server
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
export USE_SQLITE=true
.venv/bin/python manage.py migrate
.venv/bin/python manage.py createsuperuser
.venv/bin/python manage.py runserverThis is for development. Use the Compose stack for anything real.
Free Vigil answers "is my stuff up"; Business Vigil answers "can I prove to a third party that their stuff is up." Monitoring, alerting, and unlimited agents, hosts, and retention are free forever. The Business tier adds accountability features: sites, an audit log, roles, and branding. Nothing ever blocks: an expired license switches Business features off and monitoring carries on untouched. See docs/EDITIONS.md for the full feature matrix.
| Reference manual | Install options, environment variables, upgrading, backups, every task action, the API |
| Agent privileges | What each agent mode can and cannot do |
| Remote access | Checking in from off the LAN, reverse proxy, Cloudflare Tunnel |
| Reprovisioning | Rebuilding hosts from an image library |
| Roles and permissions | Who can do what |
| Security policy | Reporting a vulnerability |
| Contributing | Inbound license, sign-off, and how to send a change |
AGPLv3 for everything except server/apps_business/, which is source-visible
under a commercial license (server/apps_business/LICENSE) and requires a
Vigil Business subscription for production use. See LICENSE.