Anlora Message Guard: Block Contact Details Before a Message Is Sent

Why check your own team's messages before they reach a fan?

To stop chatters sharing contact details on OnlyFans, check each message before it reaches the fan, because one email address, phone number, link or wallet typed into a chat is already out by the time anyone could review it. Anlora's Message Guard stops a person on the team handing a fan any of those by mistake, and makes it harder for someone who tries on purpose.

In an Anlora agency the AI chats with every paying fan on its own, but people still write: when they take over one fan, send a mass message or edit the profile text. A mass message can reach many fans at once, so Message Guard checks it before it is queued, the same way it checks a one-to-one message.

How does Anlora's Message Guard check a message?

Anlora's Message Guard checks a message after a person presses send and before the message is passed to OnlyFans, in six steps.

1. 01
   A person sends

   A team member sends from the dashboard. Message Guard sits in the path the send takes, so a refusal costs nothing: no request has left Anlora. It reads the message text, the caption on a locked paid message and the profile's about, location and headline fields.

2. 02
   It confirms a person wrote it

   The first check confirms the sender is a real team member, not the AI or another automated sender. The AI's replies are never inspected, delayed, rewritten or blocked. That is a founder decision, not a setting.

3. 03
   It builds a cleaned-up copy

   For matching, Message Guard strips invisible characters, folds full-width and look-alike letters into plain ones, and turns written-out "at" and "dot" into @ and a full stop. A second copy keeps only the digits, so "5 5 5 - 1 2 3" reads as a run of digits. The text that is sent is never changed.

4. 04
   Each rule reads the copy it needs

   Every switched-on rule checks the copy that suits it. A crypto wallet is checked on the original text, because lowercasing destroys the capital letters an address depends on.

5. 05
   The mode decides the outcome

   In Blocking mode the send is refused. In Watching mode the hit is recorded and the message goes out, and the sender is told nothing. In Off mode nothing is checked or recorded.

6. 06
   The hit is logged

   Each hit writes one row: the rule, who sent it, which creator it was for and a masked fragment. The message itself is not stored.

What does Anlora's Message Guard block, and what does the agency control?

Anlora's Message Guard blocks six kinds of content, and the agency controls the mode, which rules are on, its own blocked words and the sites that are always allowed.

The agency sees all of this on the Guard tab of its team settings. It shows three mode cards (Blocking, Watching, Off), a count of hits by rule for the last seven days, a checkbox for each rule, a Blocked words list with a counter out of 200, an Always-allowed sites list with a counter out of 50, and a Recent hits log. Managers and Owners can read the log; changing the rules is the Owner's by default. Every change to the mode or the rules goes into the activity log with the before and after.

When a message is blocked, the sender reads an inline notice in the inbox, for example "This message looks like it contains a phone number. Edit it and send again." The notice names what was found and not the matched text, so it helps the person who made a mistake without teaching anyone how to get round a rule.

• Email addresses. Including ones written out, like "name at gmail dot com". An ordinary sentence such as "check it out at onlyfans.com" is not read as an email address.
• Phone numbers. A run of ten digits or more, even when the digits are spaced out. A single price, date or age is too short to count, but all the digits in one message are read together, so a message holding, say, a date and a price can reach ten and be refused.
• Links to other sites. A domain with one of the endings on an explicit list, such as .com, .link or .xyz. onlyfans.com is always allowed, and the agency can add its own sites.
• Contacts on other apps. Telegram, WhatsApp, Snapchat and link-in-bio addresses, matched on the address itself or on an app name followed by something shaped like a username.
• Crypto wallets. Bitcoin, Ethereum, Litecoin and Monero addresses.
• The agency's blocked words. Up to 200 entries, matched as whole words, so banning "cash" leaves "cashmere" alone.

How does Anlora's Message Guard compare with OnlyMonster's Message Guard?

Anlora's Message Guard starts protected with fixed contact-detail rules, while OnlyMonster's Message Guard, by its own blog post, leans on a Red List and a Yellow List of risky words, the second of which it says it updates, and an AI layer: a new Anlora team has all six rules on in Blocking mode with no list to build first, it reads written-out and look-alike versions of an email or number, and its log keeps a masked fragment instead of the message. OnlyMonster sells a pre-send filter with the same name, so the table sets the two side by side on the same nine points.

OnlyMonster's side comes from its own blog post on restricted words, read on 8 October 2026. Where OnlyMonster is stronger, the table says so: it ships a Red List and a Yellow List, the second of which OnlyMonster says it updates, which Anlora's Message Guard does not, and a Whitelist for words. Anlora's blocked-word list starts empty, there is no word whitelist, and the detection is fixed pattern rules rather than a model. A row marked not stated may be answered in OnlyMonster's documentation, which that post links to and which was not part of the sources checked. For the wider tool comparison, see Anlora vs OnlyMonster.

How do you set up Message Guard for your agency?

Set up Anlora's Message Guard in five steps on the Guard tab of the team settings; each step ends in something you can see.

1. 01
   Open the Guard tab

   Anyone whose role includes the see-what-was-blocked permission can open it. You should see the three mode cards with the current one highlighted, and the count of hits for the last seven days.

2. 02
   Pick Blocking or Watching

   A new team starts in Blocking. A team that already existed when Message Guard launched was set to Watching, and a quiet week there is the sign it is safe to switch. The chosen card is highlighted. Changing it needs the change-what-Message-Guard-blocks permission, the Owner's by default.

3. 03
   Switch on the rules you want

   A new team has all six on. Each checkbox carries a one-line hint saying what the rule catches.

4. 04
   Add blocked words and always-allowed sites

   Type a word and add it; the counter shows how many of the 200 you have used. Add your own booking page or agency site as a bare host such as agencysite.com: a full address is rejected with a message saying so. To find words worth listing, paste a draft into the free restricted-words checker, which scans for 184 words that at least two of seven public guides list. It is a separate tool, so words it flags are not added to Message Guard for you.

5. 05
   Save and read the log

   Press Save changes and a green Saved label appears next to the button. Then read the Recent hits log for false positives: each row shows the rule, the team member, the creator, the masked fragment and whether it was blocked or would have been.

Why does Anlora's Message Guard sometimes let a contact detail through?

Anlora's Message Guard sometimes lets a contact detail through because it is tuned to avoid blocking a real message, which the product's design notes judge the more expensive mistake.

A missed contact detail leaks one detail to one fan. A wrongly blocked message stops a real reply to a paying fan in the middle of a conversation, and the notes add that the person sending cannot get it out however they rephrase, so they would assume the product is broken. Message Guard is therefore built to block wrongly less often, which means a determined person can still get a number out. It is a guardrail against the careless and the opportunistic, not a censor, and Anlora does not sell it as more.

• Phone numbers under ten digits pass. A date, a price or an age would otherwise read as a number, so the floor is ten digits; a seven-digit local number with no area code gets through.
• Links use a list of domain endings. Usernames contain dots and sentences run together, so matching any two letters after a dot would block ordinary chat several times a day. An ending that is not on the list is not read as a link.
• An app name alone does not trigger it. Words like "snap" and "signal" are ordinary English, so "add me on telegram" with no handle passes, while an app address or a name followed by a handle is caught.
• Text only. It reads message text, locked captions and three profile fields. It does not read an image or a video.
• It fails towards sending. A missing setting, a damaged row or an error inside a check all end in the message going out, because a filter that stops sending silently would cost an agency more than one leaked address.

Related jobs

If the team is moving from chatters to AI, Message Guard keeps checking the people who stay: see move an OnlyFans agency from chatters to AI. Where the AI takes the night shift and people still write by day, see replace the OnlyFans night shift. For the mass messages Message Guard also checks before they are queued, see OnlyFans mass message templates.