“Bots hammer wp-login.php with no rate limit.”
Automated scripts churn through username/password combinations every day. Vanilla WordPress welcomes every attempt equally.
AI login protection · built into Pro
Brute-force detection, auto-lockouts, AI threat analysis, hidden admin usernames, a custom login URL, and a Telegram alert you can act on in two taps. It sits alongside your firewall and malware scanner, not in place of them. Built into PressBot Pro.
AI analysis costs less than $0.01 per day. The username checks are free: see what Site Check finds.
The status quo
Bots try common usernames and leaked passwords against wp-login.php around the clock, and WordPress on its own never limits how many times they can try. Most site owners only notice after the breach — three tiny defaults make that inevitable.
“Bots hammer wp-login.php with no rate limit.”
Automated scripts churn through username/password combinations every day. Vanilla WordPress welcomes every attempt equally.
“You won’t know until it’s far too late.”
WordPress sends no alerts for failed logins. Attackers can probe for weeks before anyone sees the logs — if the logs even exist.
“Manual IP blocking is tedious work.”
Editing .htaccess, wading through security-plugin dashboards, triaging one IP at a time — it wears thin fast.
How Shield works
A five-stage pipeline runs silently every time someone touches wp-login.php. You see it only when the AI decides you should.
Success or fail, Shield records the IP, username, user agent, and timestamp. Cloudflare-aware IP detection reads CF-Connecting-IP, X-Forwarded-For, and REMOTE_ADDR so you get the real attacker — IPv6 included.
When failures from one IP cross your threshold (default: 5 / 15 min), Shield acts without waiting for you.
The IP is banned. They see a “too many failed attempts” error; legitimate users never do. The block expires after your configured window (default: 30 min).
PressBot uses the cheapest available model to classify the attack: automated script or human, credential stuffing or targeted, what the attacker likely knows, and what you should do about it.
The AI’s verdict lands in Telegram with inline buttons: Block IP (30 min), Block IP (24 hrs), View Details, Mute 1h, Dismiss. Tap once. Done. No browser required.
What an alert looks like
No raw logs. No dashboards to open. Just a quick verdict from the model, the facts that matter, and the buttons that close the loop.
5 failed logins from 104.28.208.84
Username: “admin”
5 attempts in the last 10 minutes
IP auto-locked for 30 minutes
AI: High threat. Automated credential stuffing via a Cloudflare tunnel. “admin” doesn’t match any real account here; user agent is a Python script, not a browser. Recommend extending the block to 24 hours.
One tap executes the action. No dashboard, no browser.
Hide the target
WordPress publishes administrator login names in several places by default. Once a name leaks, bots keep trying it. PressBot finds those places, closes them, and tells you when a real login is being attacked.
Site Check lists every place an administrator’s or editor’s login can be read: the users sitemap, author pages and bylines, the REST API, oEmbed and feeds, and login or password-reset messages that confirm a name exists.
One switch (off by default) turns those leaks off for visitors and makes login and password-reset messages generic. The block editor, application passwords and the MCP endpoint keep working.
WordPress can’t rename a login, so PressBot does it safely for your own account: it asks for your current password, refuses names that are easy to guess, keeps you signed in, signs out other devices and emails you.
Shield lists the usernames bots are trying and flags when a real login from the leaked list is under attack; the Telegram alert says where it leaks. Ask the agent to run check_username_exposure any time.
Move the login screen to your own address. Direct visits to wp-login.php return a 404, and logged-out visitors who open a /wp-admin/ deep link get a 404 too. Set it in Shield, or ask the agent (it asks you to confirm first).
Blocks a common way bots try many passwords in one request. Leave it on only if you use Jetpack or the old WordPress mobile app.
Not sure whether your admin name is public? We wrote up what we found on our own site: Your WordPress admin username is probably public. Ours was.
Check any site’s public usernames in seconds. Our free WordPress username checker looks at five public places from the outside. It never logs in and never keeps names.
Beyond one IP
Credential-stuffing bots spread attempts across many addresses to stay under any per-IP limit. Shield looks at the whole picture, and only interrupts you when a human needs to act.
Every hour Shield checks for a username targeted by five or more different IPs within the last hour. The Telegram alert offers Block All IPs for 1 hour or 24 hours in one tap.
An address you have logged in from needs five times as many failures before it is locked out, so a mistyped password doesn’t lock you out of your own site. The Shield page lists those addresses so you can remove them.
Routine blocked-login noise goes to the daily digest. Attacks on a real account, and high or critical Plugin Guardian findings, still reach Telegram right away. Mute 1h on any alert, plus a daily limit on interruptions (default 10).
New in 1.6 · Plugin Guardian
31 WordPress plugins shipped a silent backdoor through auto-updates on April 16, 2026. Wordfence, Sucuri, and every other signature-based scanner missed it. Plugin Guardian is the layer that would have flagged it — before the payload reached your site.
eval, base64_decode chains, new outbound HTTP, wp-config.php writes, obfuscation. Semantic red flags, not signatures.
What you get
Six working pieces, zero plugin soup. Shield was built to feel like it was in the WordPress core all along.
Configurable thresholds — max failures, time window, lockout duration. Defaults work out of the box; tweak one slider if you need to.
Attack pattern, threat level, what the attacker probably knows, recommended action. Runs on the cheapest model available — under $0.01 per day.
Block IP (30 min), Block IP (24 hrs), View Details, Mute 1h, Dismiss. Tap a button — the action executes on your site. To lift a block, ask the Telegram bot to unblock the IP (it waits for your tap to confirm).
Reads CF-Connecting-IP, X-Forwarded-For, REMOTE_ADDR. Always identifies the real attacker, never the proxy. IPv6 supported.
Login data purges after 30 days on its own. No database bloat, no cron jobs to babysit, no migration headaches when you move hosts.
Ask the agent for a security audit and get a live checklist: file permissions, debug mode, SSL, database prefix, admin users, XML-RPC, and more.
Your thresholds
Configure Shield in PressBot → Settings → PressBot Shield. The defaults below are what ships, and they’re right for almost every site — change only what you know you need. Quiet Shield lives on the Shields page (home) — Shield Settings first, Quiet Shield next in the same save form — with a compact twin in Settings.
Pro · with Cloudflare
If your site is on Cloudflare, Edge Shield carries Shield’s lockouts out to Cloudflare.
Setup is one scoped Cloudflare token. Cloudflare’s free plan covers it. Cloudflare and PressBot: setup, free and Pro.
Free entry point
Site Check is in the free plugin. It looks for the traces a hacked site usually leaves: recently created administrators, WordPress core and plugin files that don’t match WordPress.org’s checksums, PHP files in uploads, unexpected must-use plugins, and scheduled tasks running code from odd places. It shows the evidence for each finding. It detects; it does not remove anything, and it is not a full malware scan. Pro re-checks weekly and alerts you on Telegram when a new warning sign appears.
Also protects your AI keys. Key Relay keeps your Claude key in your own Cloudflare account instead of WordPress, and Key Freeze locks every stored key away until you need it. How PressBot protects your site and your AI keys.
Shield is just the start
Shield lives inside the broader admin agent — content tools, WooCommerce, plugin management, AI images, Telegram, scheduled automations, and MCP for your IDE. One purchase, one toolset, every surface.
Pricing
No per-message fees. Bring your own Anthropic, Gemini, or OpenAI key for Shield, pay the provider directly — we don’t sit in the middle. DeepSeek is also available on the public chatbot.
Free
$0 · forever
Chatbot, Site Check, username protection
Pro · 1 site
$29 /yr
Shield + full admin agent
14-day money-back guarantee
Business · 5 sites
$59 /yr
Same features, more sites
14-day money-back guarantee
Agency · 50 sites
$99 /yr
Scale to all your clients
14-day money-back guarantee
Frequently asked
Shield records every login attempt and locks an IP out after repeated failures within your configured window (default: 5 failures in 15 minutes, locked for 30 minutes). The AI then classifies the pattern, such as an automated script or credential stuffing, and you get a Telegram alert with that verdict and one-tap buttons to block the IP for longer.
Largely, yes. Site Check (free) shows where your administrator and editor logins are public: the users sitemap, author pages and bylines, the REST API, oEmbed and feeds, and login or password-reset messages. The Hide usernames switch (free, off by default) turns those leaks off for visitors, and Change my username safely renames your own login. With Pro, Shield also lists the usernames bots are trying and tells you when a real login is under attack. These steps close the common leaks; they do not make a site unhackable.
Yes, in Pro. Set a custom login URL in Shield, or ask the admin agent, which asks you to confirm first. Direct visits to wp-login.php then return a 404, and logged-out visitors who open a /wp-admin/ link also get a 404, so log in at your custom address first. Setting an empty address restores the standard login.
It is built not to. Shield only locks an IP after repeated failed attempts within your window, and successful logins are never counted. An address you have logged in from needs five times as many failures before it is locked, and a single mistyped or made-up username does not lock anyone out at once. If a real user is locked out, you can unblock them from the admin, or ask the Telegram bot to unblock the IP (it waits for your tap to confirm).
No, and it is not meant to. Shield is not a firewall, a malware scanner or two-factor login. It covers the login surface and complements those tools. Site Check is a detection check for signs of a hack, not a full malware scan.
Alerts for brute-force bursts, distributed attacks, new user registrations and high or critical Plugin Guardian findings, plus a daily digest. Quiet Shield sends routine blocked-login noise to the digest and interrupts you about login attacks immediately only when a real account is under attack, and about plugins only for high or critical findings. You can mute alerts for an hour and set a daily limit on interruptions (default 10).
Free: Site Check, the Hide usernames switch, Change my username and the XML-RPC off switch. Pro (all tiers, no feature differences): Shield login tracking and lockouts, AI threat analysis, distributed attack detection, Quiet Shield, Telegram alerts and one-tap blocking, the targeted-username view, the custom login URL, and Plugin Guardian.
Shield is one part of how PressBot protects your site and your AI keys. See the security page, or compare Pro and the other plans.
Shield your login tonight
AI-powered login protection, auto-lockouts, and one-tap Telegram response — bundled into every Pro plan, for less than a coffee a month.