Support for online communities
Don't just moderate. Support.
Shepherd helps online communities turn difficult moments into opportunities for care, growth, and better conversations.
Built for Discord servers and creator communities that want safety without shame. Join the public server to follow pilot updates and try Shepherd in a real community space.
Watch the intro
Private support for online communities.
A short look at how Shepherd helps difficult moments become quieter, more supportive next steps.
How it works
Careful intervention, without hall monitor energy.
Detects moments that matterRecognizes moments like “Everyone hates me,” “I can’t stop,” or cruel pile-on language in configured spaces.
Opt-in scanning looks for signs of tension, harmful patterns, or distress in configured community spaces. Admins and creators control where Shepherd is active, which channels or content are enabled, and how sensitive the system should be.
Offers optional supportTurns a hard moment into a quiet invitation: “Want support for the next step?”
Shepherd reaches out privately where possible with a simple invitation, not a public callout. The tone is supportive: “I noticed this might be a hard moment. Want support?”
Guides better outcomesHelps someone pause before they spiral, lash out, disappear, or keep repeating the same pattern.
Users can choose faith-based encouragement or practical support paths, continue the current step, choose another track, or stop. Shepherd tracks aggregate outcomes so the experience can improve without exposing private conversations.
Trust & safety
Supportive by design. Safety-aware by default.
Admin enabled
Shepherd only operates where admins intentionally install and enable it. Admins and creators choose where Shepherd is active and configure behavior around channels, sensitivity, and available flows. Shepherd requests expected platform permissions up front, then controls behavior through feature flags.
No public shaming
Support is designed to be private, respectful, and optional. Shepherd avoids callouts, pile-ons, and hall monitor energy. When a concerning moment appears, the invitation feels like: “I noticed this might be a hard moment. Want support?”
User controlled
Users can choose practical support or optional faith-based guidance. Users can start support, continue an active step, choose another track, or stop. Faith-based guidance is available for users who want it, while practical support remains available and never second-class.
Safety first
Shepherd is not therapy, crisis care, or a replacement for trusted adults. For serious self-harm, despair, or crisis flows, Shepherd encourages immediate safety steps, trusted adults, and appropriate resources. It does not diagnose, promise outcomes, or rely only on spiritual content.
Private data
Analytics improve the product without exposing users. Public dashboards stay aggregate-only. Shepherd does not expose user-level data, identifiable community data, raw messages, private conversations, or sensitive logs.
Analytics preview
Learn what messages, flows, and moments are actually helping.
All analytics are aggregate and privacy-aware.No private conversations are exposed publicly.
Future plan
A phased path toward trusted human support.
Software-first foundation
Bot plus guided support paths, strong UX, analytics feedback, and no dependency on human availability.
Curated human network
Invite-only mentors, manual vetting, training, clear guidelines, and supervised interactions.
Scaled support network
Reputation signals, feedback loops, specialization tags, onboarding systems, and availability scheduling.
Network stewardship
Long-term exploration of transparent standards, shared accountability, mentor validation, and mission-aligned oversight.
Drill deeper
Short answers first. Details when you want them.
Faith > Gospel > Follow-up
Faith-based guidance is optional, practical support remains available, and deeper spiritual next steps are invited rather than forced. Every faith-based path keeps a Gospel gateway available from faith steps, depth modules, track completion, and support entry screens. If a user chooses to respond, Shepherd creates a durable spiritual milestone record instead of relying only on analytics events.
Data + privacy
Public analytics stay aggregate; sensitive records stay private, auditable, and access-controlled. Shepherd collects only what is necessary to provide support, track progress, and improve the system. It does not sell user data, expose private conversations publicly, or publish user-level, community-identifiable, raw-message, or sensitive-log data.
Mentor network
Human support is a later, opt-in layer with vetting, training, auditability, and strict safety boundaries. The mentor network starts with a small curated group before scaling. It avoids unsafe unsupervised contact, especially for minors, and keeps Shepherd present in the flow when handoff features are introduced.
Platform permissions
Request reasonably expected platform permissions up front, then control behavior through feature flags. The goal is to avoid re-invites, OAuth churn, and broken features in already-connected communities. Message scanning, interventions, Gospel gateway, mentor network, realtime chat, and companion mode are gated behind feature flags.
Flow engine
Clients render configured journeys; they do not hardcode Shepherd support flows. Track definitions, intake questions, branching rules, support steps, depth modules, Gospel gateway copy, CTAs, escalation messages, mentor prompts, localization strings, and feature flags come from backend config or database-backed flow versions.
FAQ
Questions skeptical admins should ask first.
What is Shepherd?
Shepherd is a support-first companion for online communities that helps respond to difficult moments with care instead of punishment. When tension, harmful patterns, or emotional struggles show up, Shepherd offers users private, optional guidance to help them pause, reset, and move forward.
Does Shepherd monitor everything?
No. Shepherd only operates where admins or creators explicitly connect and configure it, and only in channels or content where it is enabled. It looks for patterns, not opportunities to publicly call out individuals.
What happens to the messages and data Shepherd sees?
It stays private. Anything Shepherd processes is kept in secured, access-controlled databases reachable only through Shepherd's own server-side systems — never shown to other users, never posted in a public channel, and never sent to our analytics, which receive aggregate numbers only, never message text or anything that identifies a person. Private support conversations stay private, and Shepherd never publicly posts its read of a message.
What support does Shepherd offer?
Users can choose faith-based guidance rooted in a Christian perspective or practical support with simple, step-by-step tools for handling real-life situations.
Does Shepherd favor one denomination?
No. Shepherd holds one core — Jesus Christ is God and Lord, and salvation is in Him — and stays open-handed on the secondary things Christians differ over, like baptism, spiritual gifts, church structure, or end-times views. It won't push any tradition's position or quietly favor one; where something is genuinely disputed, it says so and points back to your own church. No ministry using Shepherd can make it take a side.
Can Shepherd replace therapy?
No. Shepherd is not a replacement for therapy, medical care, or crisis services. In serious situations, it encourages trusted adults, appropriate resources, and immediate safety steps.
How is Shepherd different?
Most moderation tools focus on filtering, flagging, and punishing. Shepherd focuses on support, de-escalation, and growth. Less bans. More breakthroughs
Will users be called out?
Never. Shepherd is designed to avoid shame, escalation, and public exposure. Support is offered privately or through a quiet support link when platform capabilities require a public reply.
Does Shepherd use my chats or data to train AI?
No. Shepherd doesn't train any AI on your messages — it uses an existing model through a provider whose policy is not to train on API data, and it sends every request with retention turned off so they aren't stored. Nothing you say builds or improves a model, there's no data-sharing opt-in, and Shepherd never fine-tunes on user content. It uses AI to offer support, not to harvest chats.
Is Shepherd trying to push religion on users?
No. Faith-based guidance is optional and available for those who want it. Practical support is always available, and users choose their own path.
Will real humans be involved?
Shepherd is building toward a trusted network of mentors, encouragers, and intercessors. That rollout is careful, opt-in, and governed by strong safety boundaries.
Ready to test the message?