Firewall

Every app gets its own firewall.

Attacks, scanners, fake bots and floods are stopped at LiveHQ's edge, before they reach your server. Write rules in plain English, try them in log-only first, and watch every decision as it happens.

  • On every plan, Free included
  • Blocked traffic doesn't count toward your bandwidth
  • Checked at the edge, before your server

Try it

Watch the edge decide.

Every request gets its decision at LiveHQ's edge. Turn on attack mode, switch attack signatures to log-only, or start an attack and watch automatic attack mode answer it.

shop.example.com · FirewallSimulation
Protected
Requests a second12
Stopped at the edge0
Stopped, on your plan0 bytes

Automatic attack mode is ready. It turns on by itself when traffic passes 100 requests a second and 8 times normal.

  1. 21:04:04GET /products/lanternGooglebot · 198.51.100.23BlockedFake bot · not one of Google's addresses
  2. 21:04:04GET /products/lanternChrome · BrazilPassedTo your app
  3. 21:04:03GET /products?page=12python-requests/2.32 · 192.0.2.18ChallengedBot check · HTTP library
  4. 21:04:02GET /search?q=1' union select email, password from users--Mozilla/5.0 · 203.0.113.24BlockedSQL injection · sqli:union-select
  5. 21:04:02GET /pricingFirefox · SpainPassedTo your app
  6. 21:04:01GET /blog/first-lightGooglebot · Google's addressesVerified botSearch engine · checked against its list
  7. 21:04:00GET /Firefox · SpainPassedTo your app

A simulation in your browser, sped up; no request leaves this page. This app has turned on the Protect sign-in and Let webhooks in templates, vulnerability scanners, fake bots, the bot check, attack signatures and automatic attack mode.

The console

This is the real Firewall tab.

Screenshots of LiveHQ's console as it ships, shown with example data. Every app you deploy has this tab.

In order

How a request is checked.

Every request to your app takes this path through LiveHQ's edge. The first step that refuses it ends the trip, so nothing it carries reaches your server.

Trusted IPs, allow rules and skip rules can let known traffic past later steps. Every refusal is counted and named in your Firewall tab.

  1. A request arrives at LiveHQ's edge
  2. Blocked and trusted IPs

    Trusted addresses skip the firewall. Blocked ones get 403, each with a note and an optional expiry.

    403
  3. Persistent blocks

    A visitor a rule penalized keeps getting that answer on every path, for 1 minute to 24 hours.

    Same answer
  4. Flood guard

    A ceiling per visitor for this app, far above what people produce.

    429
  5. Your rules

    Top to bottom: block, check, rate limit, allow, skip, redirect or log.

    Your choice
  6. Managed lists

    Bad networks, fake bots, vulnerability scanners and AI crawlers, if you turn them on.

    403
  7. Attack signatures

    Seven families of attacks, looked for in the path, query, headers, cookies and body.

    Log · check · 403
  8. Bot check

    Requests that look automated get the browser check.

    Log · check
  9. Attack mode

    When it's on, set by you or by itself, every visitor passes the browser check first.

    Check
  10. Leaked passwords

    A sign-in with a password known from breaches reaches your app flagged.

    Header
  11. Your appThen its access checks, and down your server's own tunnel.

What's inside

Everything it does. On every plan.

Attack signatures

It reads every request for attacks.

In the spirit of the OWASP Core Rule Set, the edge looks for seven families of attacks in the path, the query, cookies, headers and the first 16 KB of form, JSON, text and multipart bodies.

  • Log, ask for a browser check, or block with 403
  • Three sensitivities, from almost never wrong to broad
  • Every event names the signature, like sqli:union-select
  • Padding can't hide an attack: oversized values count as a detection
  • Password fields are never inspected, and the check costs microseconds
Inspected at the edgeExample
GET /search?q=1' union select email, password from users--
User-Agent: Mozilla/5.0
Cookie: cart=3
Blocked · 403SQL injection · sqli:union-select
  • SQL injection
  • Cross-site scripting
  • Path traversal
  • Command and template injection
  • Log4Shell, Spring and OGNL
  • PHP injection
  • SSRF
ModeOffLog onlyBrowser checkBlock
SensitivityLowMediumHigh

Bots

Real bots in. Fake ones out.

LiveHQ recognizes well-behaved crawlers and services by the address lists their operators publish, fetched from the official sources every 6 hours.

  • Verified: Google, Bing, OpenAI, Perplexity and Apple crawlers, Stripe and GitHub webhooks, UptimeRobot
  • Verified bots skip the bot check and attack mode; your blocks and rules still apply
  • A scraper calling itself Googlebot from anywhere else is refused
  • The bot check sends HTTP libraries, headless browsers and browsers missing the headers real ones send to a browser check
  • AI training crawlers can be turned away with one switch
Who's askingExample
  • Googlebotfrom Google's published addressesVerified
  • Googlebotfrom a rented serverRefused · fake bot
  • python-requests/2.32an HTTP libraryBrowser check
  • HeadlessChromea headless browserBrowser check
  • Stripefrom Stripe's webhook addressesVerified

Address lists from Google, Bing, OpenAI, Perplexity, Apple, Stripe, GitHub and UptimeRobot, refreshed every 6 hours.

Attack mode

One switch. Or none at all.

Turn on attack mode and every visitor passes a quick browser check before reaching your app: well under a second on a laptop, a second or two on an older phone, with nothing to click. Scripts, scanners and simple bots don't pass it.

  • Automatic attack mode turns itself on for 15 minutes when traffic jumps to at least 100 requests a second and 8 times normal
  • Each new spike starts the 15 minutes again
  • Owners and admins get an email, at most one of each kind every 6 hours per app
  • Verified bots, trusted IPs, allow and skip rules still get through
  • A solved check lasts 15 minutes to 24 hours, your choice
Requests a secondExample
Checking your browserAbout a second. Nothing to click.
Attack mode turned on for shopshop's traffic jumped far above normal, so LiveHQ turned on attack mode for it.Open firewall

Rules

Rules in plain English.

Every rule reads as a sentence. Match the path, query, method, domain, IP, country, continent, user agent, a header, a cookie, the body, verified bots or network lists, and combine conditions with OR groups.

  • Block, ask for a browser check, rate limit, allow, skip, redirect or log
  • Rate limits per visitor, visitor and path, header, cookie, path or everyone, over 10 seconds to an hour
  • Keep refusing a visitor on every path for 1 minute to 24 hours
  • Your own block page in text, HTML or JSON, and redirects
  • Skip rules let your API and webhooks past the checks machines can't solve
Rules, top to bottomExample
  1. 1Protect sign-inEnforced

    When the path is /login or /signin and the method is POST, limit each visitor to 10 requests a minute, then block, and keep blocking that visitor for 15 minutes.

  2. 2Let webhooks inEnforced

    When the visitor comes from Stripe's webhook addresses and the path is /webhooks/stripe, or when the visitor comes from GitHub's webhook addresses and the path is /webhooks/github, skip the bot check and attack mode for the request.

  3. 3Admin from the office onlyLog only

    When the path starts with /admin and the visitor's IP is not in the Office list, block the request with your own 404 response.

Start from a template

Each one creates an ordinary rule or setting you can read and change.

  • Protect sign-in
  • Block attack payloads
  • Let webhooks in
  • Protect your API
  • Block vulnerability scanners
  • Block countries
  • Let search engines in
  • Block Tor
  • Check cloud traffic on sign-in
  • Lock a path to a list
  • Ban scrapers for an hour
  • Maintenance redirect
  • Block AI crawlers
  • Block fake bots
  • Block bad networks
  • Watch for bots
  • Flag leaked passwords

Lists for every address

Block or trust IPs and ranges, each with a note and an optional expiry. Keep named lists for the whole workspace: change one and every app that uses it follows within seconds. Network lists keep themselves fresh: Tor, AWS, Google Cloud, Oracle Cloud, DigitalOcean and Spamhaus DROP.

Leaked passwords, flagged

On sign-in, the edge checks the password against Have I Been Pwned without sending it: only the first 5 characters of its SHA-1 hash leave the edge. A breached password reaches your app with a header, so it can ask for a reset or a second factor. Nothing is refused.

Try before you enforce

Every rule and every protection that refuses requests can run in log-only mode first, counting what it would have done. Edit a draft, review the changes, then publish. The last 25 versions are kept and any can be restored. Export to JSON, import, or copy another app's firewall.

See every decision

The last hour, 6 hours or 24 hours: requests, blocks, rate limits, browser checks and logs; the top IPs, countries, paths, causes and user agents; and recent events with their details.

Fair by design

Security isn't a paid extra.

What each plan holds
SizeFreePro
Rules per app10100
Blocked and trusted IPs per app1002,000
Shared workspace lists350
Addresses per shared list1,00010,000

Every feature is the same on both. See the plans

What it doesn't do.

A firewall you can trust tells you where it stops.

  • The browser check stops scripts, scanners and simple bots, and makes floods expensive. It doesn't stop a person, or a real browser driven by an attacker who hides automation well.
  • Automatic attack mode reacts within about 10 seconds, so the first seconds of a flood still reach your server. The flood guard and your rate limits cover those.
  • It can't tell a launch-day rush from an attack. People pass the check; API clients and webhooks need a skip rule while it's on.
  • No machine-learned attack or bot scores, no TLS fingerprints and no rules by network operator (ASN). LiveHQ uses signatures, header checks, published address lists and the browser check instead.
  • Only the first 16 KB of a request body is read. Compressed bodies and responses aren't inspected.
  • History covers the last 24 hours, by design.
  • Cloudflare's network DDoS protection sits in front of LiveHQ's edge. This firewall works on the requests that reach it.

Questions

Firewall questions, plainly.

Does it slow my app down?

Barely. The edge checks each request as it passes, and attack signatures cost microseconds. Only two things can add a pause, and only when you turn them on: the browser check, which a visitor solves once per clearance, and the leaked-password lookup on sign-in, which waits at most a second and is cached for an hour.

Will it block Google?

No. Googlebot, Bingbot and the other verified crawlers are recognized by the addresses their operators publish, and they skip the bot check and attack mode's browser check. Your own blocks and rules still apply to them. A scraper that only calls itself Googlebot is refused once you turn on fake-bot blocking.

What about my API, mobile app and webhooks?

Machine clients can't solve a browser check. Stripe and GitHub webhooks and UptimeRobot are verified by their published addresses and pass attack mode. For your own API and apps, add a skip rule for the bot check and attack mode on their paths; rate limits and attack signatures still apply. The Let webhooks in template sets this up for Stripe and GitHub.

Does blocked traffic count toward my bandwidth?

No. The edge answers blocks, rate limits, browser checks and redirects itself, before the bandwidth meter, and never forwards a refused request's body. LiveHQ pays for that traffic, not you.

Is the firewall on the Free plan?

Yes, every feature. Plans only size the lists: Free has 10 rules and 100 blocked or trusted IPs per app, Pro has 100 and 2,000.

How do I start safely?

In log-only mode. Every rule and every protection that refuses requests can count what it would have done before it acts, and the Block attack payloads template starts attack signatures that way. Read the events, then switch to enforcing. If something goes wrong, restore an earlier version.

What does a blocked visitor see?

A short LiveHQ page or a plain answer with its status, or your own response if you wrote one. Nothing is blocked silently: every refusal names its cause and shows up in your Firewall tab.

Does it protect private apps?

Yes. Private apps run the firewall before their sign-in check, so the sign-in flow is protected too.

There's no point knocking on a door that never opens.

Let them knock. Sleep anyway.

Start free, deploy an app and open its Firewall tab. Every protection is there on every plan, ready to try in log-only first.