One subscription per server

The free plan blocks every scam the shared network already knows, for as long as you want and at no cost. Paid plans add the part that costs money to run: reading images nobody has seen before, and acting on them without waiting for a moderator to wake up. Cancel any time from the Stripe portal.

Free

€0 free forever

Blocks scams the network already knows. Costs you nothing.

  • Unlimited blocks from the shared scam network
  • No AI analysis of new images
  • Reports in the channel you choose
  • 7 days of history
Start free

Guard

€4.99 per month, per server

Reads new images with AI and acts on its own.

  • Everything in Free
  • 2,000 AI analyses of never-before-seen images
  • Automatic ban, kick, timeout or delete
  • Shared network blocks known scammers on sight
  • 30 days of history with archived evidence
  • Reverse any action in one click
Get Guard

Pro

€6.99 per month, per server

Protection and an assistant. Run either one, or both.

  • Everything in Guard
  • 4,000 AI analyses a month
  • 90 days of history
  • CSV export of every action
  • Webhook alerts into your own tools
  • Check single images, not just pairs
  • FAQ assistant: 1,500 answers a month from your own knowledge base
  • Run the assistant on its own if that's all you need
Get Pro

Network

€9.99 per month, per server

High volume, sharper checks, and a busier assistant.

  • Everything in Pro
  • 7,000 AI analyses a month
  • Stronger model re-checks every borderline call
  • 12 months of history
  • Read API with your own key
  • FAQ assistant with 5,000 answers a month
  • Direct support, answered by a human
Get Network

Questions we get every week

What counts as an analysis?

Only images nobody in the network has seen before. If a scam was already judged on another protected server, recognition is instant and doesn't touch your allowance. During a spam wave almost every block is free, because scammers repost the same images across dozens of servers.

What is the FAQ assistant?

On Pro and Network, members can ask questions with a trigger you choose and get an answer drawn from a knowledge base you write. You set the character it speaks in and the exact sentence it must use when it doesn't know. It never answers from anything except your own entries, so it can't invent rules, dates or mechanics that don't exist.

Why is the free plan free?

Because it costs us nothing to run. Blocking a scam we've already identified is a database lookup. Reading a brand-new image costs real money every time, which is what you're paying for. No trial timer, no bait: the free plan stays free forever.

What happens when I run out of analyses?

Shepherd keeps blocking everything it recognises by fingerprint or through the shared network, and only stops analysing brand-new images until the month rolls over. Your quota is a ceiling, not a meter — there are no surprise charges.

Does the bot wait for me to approve a ban?

Never. On a paid plan Shepherd bans, clears the message and then posts the evidence. The buttons in that report are there to undo a mistake after the fact, not to authorise anything. A scam posted at 3am is gone at 3am.

What exactly does the free plan not do?

Two things. It never reads a new image, so a scam that hasn't been caught anywhere else yet goes through. And it never acts: no bans, kicks, timeouts or deletions. What it does do, without limits, is block anything the shared network already recognises and tell you about it.

Do I need my own OpenAI account?

No. The analysis is included in the price. No API key, no second subscription, nothing to set up.

I run several servers, is there a bundle?

Each server has its own subscription. If you run a network of communities and want to talk about it, email us at filippocastagna44@gmail.com.

What if it bans the wrong person?

You reverse it from the log channel or the dashboard. The member is unbanned and that image leaves the shared network. If you'd rather not take the risk at all, leave Shepherd in watch-only mode: it reports and never acts.

Which permissions does the bot need?

View channels, read message history, manage messages, kick, ban, timeout, and send messages with embeds. No Administrator. A moderation bot should never need it, and you shouldn't trust one that asks.

Try it on your busiest server

The free plan is enough to find out whether you have a problem.

Add to Discord