Skip to content

Moderation where every decision has a reason and a record.

Automated checks sort reports on text, images, audio and video. Reviewers decide the hard cases, users can appeal, and each decision is logged with the rule it applied.

Sentinel
KettleyardTrust and Safety
Queue24AppealsAccountsPolicies
Search/
Queue2417 over SLA
MineTeam
  • Kitchen table income+02:14Spam14
  • Account closure scam01:12Scam9
  • Insults to a referee11:48Harassment7
  • Fake support account28:55Impersonation4
  • Log off and stay off17:20Harassment3
  • Pile on from C-114NewHarassment5
38
Wellbeing

min exposure today

Break in 12 min

Blur and greyscale on

SN-48207In reviewRemoved

Targeted insults aimed at a volunteer match official

§ 4.2Targeted insults7 reports, 2 from trusted flaggers
11:48SLA left
Eastside Sunday LeagueMatch day threadkettleyard.com
@tess_vossReferee

Results from Sunday are up. Thanks, everyone, for keeping it friendly.

Eastside 2, Brickfield 1
4831
Q
@quietbrook_77New account

@tess_voss you are honestly the most clueless ref this league has ever had1. Everyone knows you get paid off by the Brickfield side2. Don’t bother turning up on Sunday, we will all be waiting for you3. The full list of your “mistakes”: refwatch.kettle.link/tv4

Reported 7 timesReply hidden
Attachments3
HHold to reveal
Hash match
Classifier scores90% interval
1harassment.targeted
0.88§ 4.2 a

Under the automatic line, so a person decides

Highlighted in the reply
Span 1
Removed automatically from
0.92
Similar replies removed
94%
2harassment.accusation
0.74§ 4.2 b
3violence.threat
0.41§ 3.1
4link.offplatform
0.73§ 8.3
Decision
5 replies hidden across C-114
RRemoveRemovedSuggested
AApprove
EEscalate
WWarn
  • Reply hidden
  • Strike 2 of 3
  • Author can appeal
Undo for 10 min
Q@quietbrook_77Account82Risk
Joined
19 days ago
Reports upheld
11 of 23
Strikes
1 of 32 of 3
Device
Shared by 4
Linked accounts13

Replies within 9 minutes

  • @quietbrook_77This case
  • @kettle_ultraSame device
  • @brick_beeSame device
  • @touchline9Joined same day
  • @sundaybootsSame device
Cluster C-1145 accounts
Throughput

Harassment queue, last 24 h

Inflow
3,355
Decisions
3,221
Breached SLA
41
DecisionsInflowp90 to decision

SentinelA trust and safety queue with policy scores and appeals. Fictional business, built to show the work.

What to look at

What this covers

Every service in this family.

Policy-driven review and incident tools for online communities and user-generated content. 16 services

  • Chat and community moderationMessage review, queues and authorized moderation actions.
  • Spam and abuse controlsRules and signals for unwanted or abusive activity.
  • Image, video and audio reviewMedia review pipelines and policy-specific escalation.
  • Marketplace listing moderationListings, seller content and product-policy checks.
  • Reviews and commentsReview submitted feedback and discussion contributions.
  • Member verification and admissionParticipation rules, verification workflows and access gates.
  • Moderator dashboardsQueues, incidents, evidence and authorized tools.
  • Appeals and false-positive reviewReview disputed decisions and proposed corrections.
  • Incident and evidence managementCase timelines, retained evidence and review records.
  • Policy configurationCommunity-specific rules, thresholds and escalation paths.
  • Audit and accountability recordsWho acted, what changed and the supporting evidence.
  • Platform integrationsSupported Matrix, forum, community and social-platform APIs.
  • Rate limits and coordinated-abuse signalsDetect suspicious activity patterns within agreed data boundaries.
  • Attachment review workflowsScan supported uploads and route uncertain results for review.
  • Human review and escalationClear review paths for consequential or uncertain decisions.
  • Moderation analyticsMeasure activity, appeals, errors and policy outcomes.
Technology options

Services in TypeScript or Python, with Go or Rust where justified, connected to platform APIs, queues and databases. React dashboards and model-assisted classification can support the reviewers. Every consequential action is authorized on the server.

How we choose technology

Questions

Can it work with our existing platform?

Yes. The engine connects to your product through its own interface, so content is checked where it is posted.

Does AI make the final decision?

Only where you decide it should. Clear cases can be automated; anything uncertain goes to a person, and an undo path is part of the design.

Which kinds of content can it check?

Text, images, audio, video, files and chats, each with checks suited to it.

How are reviewers protected?

Blurring, content warnings and workload limits are built into the queue.

Start

Tell us what you are building.

A short description is enough. You get a straight read on scope, cost, and our honest take on the best approach, before anyone commits.

Write to [email protected] · New projects and general questions

  • Built from zero, no templates
  • Hosted and maintained by the same team
  • Scope and cost agreed before we start
  • Updates, monitoring and support after launch
  • Findable in search and AI answers