Sentinel

Stops attacks on your app, and says exactly why.

Sentinel runs in front of your web app, API or AI endpoint. It checks every request for attacks and abuse, and every decision comes with the rule, the field and the text that set it off. It runs on your own server, and the free edition covers any number of them.

What Sentinel prints in your terminal. The dashboard shows the same, with an Allow button for anything that shouldn't have been stopped.

What happens to a request

Cheap checks run first, so a banned address is turned away before any pattern is matched. You start in monitor mode: everything is checked and logged, nothing is blocked, until you've seen what it would do on your traffic.

  1. Who is asking

    Banned IPs, countries you don't serve, honeypot paths only scanners visit, known attack tools.

  2. How often

    Rate limits per path, password-spraying detection, an optional proof-of-work check for bots.

  3. What it contains

    Sentinel's rules after decoding evasions, the OWASP Core Rule Set, and an ML model for attacks no rule describes.

  4. What goes back

    Database errors, stack traces and keys in responses are withheld. Card numbers are masked.

Tested against real traffic

On open-appsec's public WAF comparison: real browsing traffic recorded on hundreds of sites, and 11,054 public attack payloads. Every WAF ran with default settings, in blocking mode, in front of the same app. Sentinel's model never saw the sites counted below.

GatewayAttacks blockedNormal requests wrongly blocked
Sentinel99.9%2.5%
ModSecurity with the OWASP Core Rule Set90.3%26.7%
BunkerWeb90.3%28.8%

This dataset is full of unusual requests, so every number is worse than on a typical site. Method, the rule fixes behind these numbers and their limits: benchmarks.

Try it on your machine

Put Sentinel in front of an app on port 3000:

docker run --rm -it -p 8080:8080 -v sentinel-data:/var/lib/sentinel \
  -e UPSTREAM_URL=http://host.docker.internal:3000 \
  --add-host=host.docker.internal:host-gateway \
  ghcr.io/thefirstterminal/sentinel

Open the one-time link it prints to create your admin account. On a Linux server, one line installs it as a service; there are also Debian and RPM packages, Homebrew, Helm and plain binaries. Every way to install.

Free and Pro

Sentinel Free

$0 any number of servers, in production too
  • Rules, OWASP Core Rule Set and the ML model
  • Prompt-injection guard for AI endpoints
  • Rate limits, bans, honeypots, bot check
  • Several sites, load balancing, automatic HTTPS
  • Dashboard with 2FA, alerts, syslog and OpenTelemetry

Sentinel Pro

$29 per server a month, or $290 a year
  • Everything in Free
  • Single sign-on and 2FA logins for your own apps
  • Access rules by role, department, ownership, time and network
  • Step-up checks for unusual countries, IPs and hours
  • Client certificates, tamper-evident audit log, email support

The first 50 customers pay $19 a month for as long as they renew. Write to support@tft.markets with your company and number of servers.