How Guardiobot works.

Guardiobot is added through the server owner dashboard. You need Administrator permissions on the target server.

1. Log in with Discord — go to the dashboard login and authorise with Discord OAuth2.
2. Select your server — the dashboard lists all servers where you have Administrator permissions.
3. Invite the bot — click Add Guardiobot. Discord will ask you to confirm the required permissions.
4. Configure settings — set your log channel, verification channel, and feature toggles. The bot works with defaults immediately, but reviewing these is recommended.

The bot starts working on four things the moment it joins — no additional setup required for these to be active:

When a new member joins, Guardiobot sends them a DM with a 6-digit code. They enter it in your verification channel. Then they answer a short maths challenge. Both steps must pass before the verified role is granted.

This two-factor approach stops most automated bot waves without being a barrier to legitimate users. If a member doesn't receive the DM, they can request the code again in the verification channel.

These slash commands are available to server administrators and moderators directly in Discord:

1. Right-click the message (or long-press on mobile) to open Discord's message action menu.
2. Select "Apps" from the menu.
3. Click "Report to Guardiobot" — a modal opens asking you to choose a category.
4. Select a category and submit. You'll receive confirmation that the report was received.

Every report enters a queue reviewed by Trust & Safety volunteers — not a bot. A volunteer claims the report, reviews the message and context, and takes one of these actions: warn, restrict, escalate to senior staff, or issue a removal. Enforcement decisions apply across the entire Guardiobot network.

• Immediate — you receive confirmation your report was received.
• Within 24 hours — a T&S volunteer claims and reviews the report.
• After action — the decision is recorded in the audit log. You won't receive the specific outcome, but actions are logged and reviewable.

Do not submit duplicate reports. Multiple reports for the same content don't speed up review — they create extra work for volunteers.

Guardiobot maintains a safety classification for users across the network based on verified offence history:

Class A and B users may be automatically blocked from joining servers depending on each server's configuration. If you believe you were classified incorrectly, you can submit an appeal through the server owner dashboard.

For content involving immediate danger, illegal activity, or child sexual abuse material (CSAM):

• Report through the normal Guardiobot flow — CSAM and illegal content are highest priority and reviewed within 1 hour.
• Email abuse@guardiobot.org for urgent cases that need direct escalation.
• If someone is in immediate physical danger, contact emergency services in your country first.
• CSAM is reported by Guardiobot to the National Center for Missing & Exploited Children (NCMEC) in accordance with legal obligations.

For crisis lines and safety organisations by country, see our Emergency Resources page.

Every message goes through the following steps in sequence:

1. Spam detection — duplicate messages sent rapidly are caught by an in-memory sliding-window check. No external calls, no latency. Match → delete + 5-minute mute.
2. Content detection engine — non-spam messages are scored 1–10 using four methods: keyword matching, regex patterns, SHA-256 content hashes, and URL pattern matching. Runs locally, no external API dependency.
3. Adaptive learning sample — a percentage of messages (configurable: 1%, 10%, or 50%) updates the community's behaviour model asynchronously. Doesn't affect detection speed.
4. Severity-based action — score 8–10: automatic deletion + enforcement. Score 5–7: report created for volunteer review. Score 1–4: logged only.

Each server develops its own baseline behaviour profile over time. The system tracks five signals: message length, emoji frequency, link frequency, capitalisation rate, and mention frequency. As your server's normal posting style changes — events, staff changes, rebrands — the detection thresholds adjust, reducing false positives over time.

Server-specific overrides are also stored: words that are normal in your community can be whitelisted; terms that are particularly sensitive can be added to a custom blacklist. Both are managed through your dashboard.

The anti-raid system monitors join events in real time and responds to coordinated attacks automatically.

• Join floods — abnormally high join rates trigger rate limiting and optional automatic lockdown.
• VPN/proxy detection — joining accounts using known VPN or proxy IPs are flagged. Requires third-party API keys (ProxyCheck, IPHub, or VPNAPI) configured in your dashboard.
• Alt-account detection — accounts with patterns consistent with alt behaviour are flagged for review.
• Account age enforcement — accounts created too recently can be automatically kicked or restricted. Threshold is configurable per server.

Lockdown mode can be triggered automatically when a raid is detected, or manually via /lock. While in lockdown, no new joins are processed. A moderator must manually lift it via /unlock or through the dashboard.

When a user is banned in one participating server, that ban is propagated to every other participating server automatically. Each server has a trust score. High-trust bans are auto-accepted. Lower-trust bans go to a manual review queue where a volunteer approves or rejects before enforcement.

Network participation settings are managed by Guardiobot's Management team — not by individual server owners. This is intentional, to prevent servers from being added to or removed from the network without oversight.

Enforcement also applies on join: when a user with an active cross-server punishment joins any protected server, the punishment is re-applied automatically. This closes the most common evasion path — leaving a banned server and joining another in the same network.

All core protection features are free. Subscriptions unlock optional extras.

Subscriptions are managed through Stripe. To subscribe, go to the Subscription section in your server owner dashboard.

Community Manager subscribers can enable an optional AI conversation assistant in a designated channel. The bot uses the Claude AI model — but you provide your own Anthropic API key. Guardiobot does not hold a shared key and cannot access your usage or conversation logs.

1. Go to Dashboard → AI Settings for your server.
2. Enable the AI chatbot and select the channel it should respond in.
3. Enter your Anthropic API key. The key is validated on save and stored encrypted (AES-256-GCM) — it is never logged.

Billing is handled by Stripe. To view invoices, update your payment method, or cancel your subscription, go to Dashboard → Subscription → Manage Billing. This opens the Stripe Customer Portal where all billing actions are handled.

Cancellations take effect at the end of the current billing period. Your server keeps access to paid features until then. If you cancel and re-subscribe, a new billing cycle starts from the date of re-subscription.

The public API is accessible via https://dashboard.guardiobot.org. Developer tier subscribers can generate API keys from Dashboard → API Keys. Each account can hold up to 5 active keys.

Authentication uses a Bearer token passed in the request header:

Authorization: Bearer gb_<your_key>

The rate limit for API key requests is 60 requests per minute. Exceeded limits return HTTP 429. Keys can be revoked at any time from the dashboard.

The following public endpoints are available to Developer tier subscribers. All responses are JSON.

{
  "discordId": "123456789012345678",
  "classification": "C",
  "activeRestrictions": false,
  "networkBanned": false
}

All error responses include a JSON body with a message field describing the issue.