From installation to a safe first test
Sentinel’s open beta is deliberately focused: one reliable honeypot protection system, configured per Discord server, with clear reporting and no stored message content.
Install Sentinel
A server owner or administrator can add Sentinel through Discord’s installation screen. Choose the correct server, review the requested permissions, and complete authorisation.
- Open the installer. Use the button below or the installation link on the home page.
- Select a server. Discord only lists servers where you can add applications.
- Review access. Sentinel needs message and moderation permissions only for the actions you enable.
- Open the dashboard. Sign in separately so Sentinel can show servers you own or manage.
Installing adds the bot to a server. Signing in requests only your basic Discord identity and guild list. Sentinel does not request your email address.
Required permissions
Discord evaluates permissions per channel and by role hierarchy. A server-level permission can still be blocked by a channel override.
| Permission | Why Sentinel needs it |
|---|---|
| View Channel | Detect activity in selected honeypots and reach the incident channel. |
| Send Messages | Publish the persistent warning and incident reports. |
| Embed Links | Display Sentinel notices and reports in their intended format. |
| Read Message History | Update the persistent warning and process channel activity reliably. |
| Manage Channels | Create a honeypot or private incident channel when requested during guided setup. |
| Manage Messages | Delete the triggering message for enforcement modes. |
| Moderate Members | Apply the configured timeout action. |
| Ban Members | Perform a soft-ban and immediately remove the ban. |
Place Sentinel above ordinary members it may need to action. It cannot moderate the server owner, administrators, or members whose highest role is equal to or above its own.
First-time setup
Run /sentinel setup for a private, guided setup inside Discord. Sentinel can use existing channels or create a named honeypot and a private sentinel-logs channel after you review and confirm every choice.
- Choose or create the honeypot channel.
- Leave the action on Observe and report only for the first test.
- Choose a separate staff-only incident channel or let Sentinel create one.
- Review the complete configuration.
- Confirm. Only then will Sentinel create channels, save protection, and publish its warning.
After setup, use /sentinel settings to view or change the shared action and incident destination. Use /sentinel honeypot add, create, remove and list to manage channels. Adding channels preserves settings and safe testers. Rerunning setup shows the existing configuration.
You can also configure existing channels from Your servers. The free open-beta plan supports one honeypot. Clearing the dashboard selection disables honeypot protection without removing Sentinel.
How honeypots work
A honeypot is a channel legitimate members have no reason to use. Automated spam accounts often post across every visible channel, so touching the trap provides a high-confidence signal without interpreting normal conversation.
A message appears in a configured restricted channel.
Sentinel checks exemptions. Server managers and administrators do not trigger enforcement.
The configured action runs. Sentinel records the actual outcome, not merely the requested action.
A report is sent. The triggering message body is never retained.
Channel naming and access
Use a restricted name an automated account might encounter, while clearly warning real members not to interact. Examples include #restricted-access, #system-reserved, or #do-not-post.
- Keep the channel visible but clearly restricted for ordinary members.
- Do not use it for real announcements, verification, or staff conversation.
- Keep incident reports in a different, staff-only channel.
- Do not remove or contradict Sentinel’s persistent warning.
Enforcement actions
| Mode | Behaviour | Best use |
|---|---|---|
| Observe | Reports the incident without modifying the message or member. | Initial setup and testing. |
| Delete | Deletes the trigger without actioning the member. | Low-risk containment. |
| Timeout | Applies the configured timeout duration and attempts to delete the trigger. | Temporary containment. |
| Soft-ban | Bans the member, clears recent messages, then immediately unbans them. | High-confidence spam. |
If Discord blocks part of an action, Sentinel reports whether it succeeded, partially completed, or failed. A failed automatic unban is reported as partial so moderators know the member remains banned.
Test safely
- Open the server dashboard and add your test account's Discord user ID under Individual safe testers, or select a trusted safe tester role and confirm the bypass warning.
- Post a harmless message in the honeypot from that account.
- Confirm the staff report is labelled as a safe test and shows that no enforcement was attempted.
- Remove the account from the safe list only when you deliberately want to test real enforcement.
Safe testers can include server administrators, who Sentinel otherwise ignores. Their test incidents are retained under the same 30-day policy but are not included in public threat totals.
Discord role hierarchy and Sentinel’s administrator exemption make those accounts unsuitable for enforcement tests.
Incident reporting
During open beta, Sentinel sends an immediate report for every trigger. It identifies the member and channel in Discord, configured response, actual outcome, a safe failure code when relevant, and an incident ID.
Reports explain what happened without copying the triggering message into Sentinel’s database.
Noise controls
Weekly aggregate summaries, dashboard-only reporting, and immediate-plus-weekly reporting are planned as free usability choices. A weekly summary will contain totals and outcomes, not a delayed list of named incidents.
Coming soonDashboard guide
The dashboard lists only servers Discord says you own or can manage. For each installed server you can select the honeypot, enforcement action, and incident-report channel.
- Protection status
- Whether a honeypot is configured and which action is active.
- Channel limit
- How many honeypots the current entitlement allows.
- Installation required
- You can manage the server, but Sentinel is not installed there.
Recent incident history will appear here after its authorised API endpoint is ready.
Incident history coming soonPrivacy and retention
Sentinel stores only what is required to explain and operate protection:
- Guild and channel identifiers.
- A keyed pseudonymous subject reference instead of a recoverable Discord user ID.
- Configured action, actual outcome, safe failure code, reason, incident ID, and timestamp.
- No message body, attachment, Discord OAuth token, or email address.
Raw incidents are automatically deleted after 30 days. Cleanup runs at startup and every 24 hours.
Read the full privacy notice →Troubleshooting
Sentinel does not react to my test
Confirm the channel is saved, use a non-administrator account, and ensure Sentinel can view the channel. Server managers are exempt.
The dashboard cannot save protection
Check View Channel, Send Messages, Embed Links, and Read Message History. Sentinel must create or update its warning before saving.
Timeout or soft-ban failed
Move Sentinel above the test member’s role and verify Moderate Members or Ban Members. The report should show the actual result.
A server is missing from my dashboard
Sign out and back in for a fresh guild list. You must own it or hold Administrator or Manage Server.
Planned protection and controls
These are roadmap items, not active open-beta protection.
Noise controls
Weekly aggregates, dashboard-only mode, and immediate plus weekly reports.
Incident history
A 30-day view of requested actions and actual outcomes.
Setup templates
Reusable channel placement and staff-role presets for larger communities.
Link Shield
Layered local rules, Sentinel intelligence, and external reputation checks.
Spam detection
Focused coordinated image and message-spam detection.
Guided maintenance
Edit individual settings without rerunning the complete setup flow.