Your server, guarded by AI as it lands
FreeGuard watches text, images, reactions and voice across 7 detection layers — then hands your staff a beautiful OAuth-secured console to manage strikes, bans and appeals. No subscriptions. Every record encrypted.
Real-time analysis across every surface of Discord — configured per server, adjusted live, no restarts.
Every message passes the full stack. A lightning demo — watch the sweep:
The bot never babysits a config menu — it watches, scores, warns, and lets your staff decide from the console.
Messages, attachments, reactions and voice streams are analysed live against the server's ruleset.
Violations increment encrypted per-user strike records with reason, category and timestamp.
Thresholds auto-delete, DM-warn and eventually auto-ban — with automatic lift after a set duration.
Banned users press a button in their DM, submit a form, and staff approve or deny from the console.
Commands surface automatically based on Discord permissions — everyone gets self-service, staff get the guard toolkit, admins get the control surface.
Discord OAuth login. Per-guild. Warn, role, kick, ban, resolve appeals and search the live audit trail — all behind an allow-list of server IDs.
All six views pre-fetch at login, so tab switches render from cache in near-zero time — even from a Raspberry Pi.
Access requires the session user to moderate an allow-listed server. No server ID, no access.
AES-256-GCM at rest, PBKDF2 key derivation, HMAC file naming — every stored byte is protected.
# install $ npm install $ cp .env.example .env # add DISCORD_TOKEN # run bot + console $ node index.js → bot online ✓ → dashboard on :3000 → funnel → raspberrypi.tail6bc7fe.ts.net # as a service $ sudo ./install-dashboard.sh → systemd active (running) ✓
One OAuth click with Administrator permissions — or self-host your own instance from source.
Log in with Discord, add your server's ID to the allow-list, and every view is live.
Change strike thresholds, durations and systems over 30+ runtime settings — no restarts.
The full agreements, written in plain language. Everything about data, encryption, AI and your rights is spelled out below.
These Terms govern the use of the FreeGuard Discord bot and its web console ("FreeGuard"), developed independently by Adam Eyd. By inviting FreeGuard to a server, configuring it, or using the console, you accept these Terms. If you do not agree, remove the bot.
Inviting, configuring or commanding FreeGuard constitutes acceptance of these Terms. Server owners act for their community; moderators using the console act under the owner's authority. A single instance may serve multiple servers, and each server accepts these Terms independently.
The source code is released under the MIT License — you may self-host, modify and redistribute it freely. The FreeGuard name, logo, and the "FreeGuard" trademark remain the property of Adam Eyd and may not be used to brand derivative builds without written permission.
You agree not to abuse FreeGuard, including: reverse-engineering a hosted (non-self-hosted) instance, probing or tampering with the console beyond an account you control, deliberately evading bans or strikes imposed by rules you configured, or attempting to disrupt another server's instance. Violations may result in the removal of FreeGuard from your server without notice.
FreeGuard executes the rules its operators configure. Strike thresholds, bans, warnings and appeal outcomes are decisions made by each server's staff — FreeGuard is the tool, not the judge. Automated moderation, including AI scoring, is best-effort and may occasionally misjudge; server staff remain responsible for reviewing appeals and correcting outcomes. FreeGuard does not warranty that any violation will be caught.
Content may be submitted to free AI tiers — OpenRouter for text analysis and Hugging Face for image scoring and voice transcription — solely to produce moderation scores. FreeGuard retains only the configured ledger (strike, ban and log records); provider requests are transient and governed by each provider's own policies. No free tier queues, wait times or billing applies to your community.
Banned members may appeal through the button in their ban DM, a slash command, or the console queue. Approving an appeal instantly unbans the member, clears their strikes and DMs a fresh join link. Denying an appeal keeps the ban in place and records the decision in the audit trail.
All stored records are encrypted at rest with AES-256-GCM and keyed with PBKDF2 (SHA-512, high iteration) before they touch disk. Files are addressed by HMAC-SHA256 rather than raw user IDs, and audit entries are encrypted line-by-line. This applies to both public and self-hosted instances; see the Privacy Policy for full detail.
FreeGuard is provided "as is" without uptime guarantees. The public console, funnel and web console may be restarted, updated or taken offline for maintenance at any time, and FreeGuard disclaims liability for any downtime, data loss, or interrupted moderation.
To the maximum extent permitted by law, FreeGuard, its authors and Adam Eyd are not liable for indirect, incidental, consequential or punitive damages arising from the use or inability to use FreeGuard — including wrongful bans, missed violations, or harm caused by members who were not detected in time. In no event does total liability exceed the amount you paid to use FreeGuard, which is zero (free).
You may stop using FreeGuard at any time by removing the bot from your server. Data you configured is retained on the host it ran on; the operator may purge ledger data after termination in line with the Privacy Policy. FreeGuard may terminate access to its public services at its discretion, with reasonable notice where feasible.
These Terms may be updated to reflect new features, legal requirements or security practice. Material changes are announced in the changelog and community server before they take effect. Continued use after an update constitutes acceptance of the revised Terms.
Unless otherwise stated by the operator of a self-hosted instance, these Terms are governed by the laws of the jurisdiction in which the hosting operator is established. Disputes will be resolved amicably in the first instance through the support server before any formal proceedings.
FreeGuard prioritises defensive privacy: it stores the minimum needed to moderate, encrypts everything it does store, and never sells data. Because the bot is self-hostable, the "data controller" is the operator of the instance your server runs on. This policy covers both the public instance and self-hosted deployments.
FreeGuard stores only what moderation requires: Discord user and guild IDs, encrypted strike and ban records, appeal submissions, an activity ledger, and configuration values (channels, thresholds, exemption roles). Messages, images and voice streams are analysed at runtime but are not stored beyond the configured ledger. The web console receives a short-lived Discord OAuth token that expires; session tokens are never persisted to disk.
Every stored record is encrypted with AES-256-GCM before it reaches disk. Keys are derived with PBKDF2 (SHA-512, 310,000 iterations) and support versioned rotation. User files are named by HMAC-SHA256 of the ID — raw user IDs never appear in filenames. Each audit log line is encrypted independently so a breach of one file exposes nothing else. Secrets live only in a gitignored .env and are never logged or committed.
Data is never sold, rented or shared for advertising. It is used exclusively to operate moderation for your server.
Content may be transmitted to free AI tiers to produce moderation scores: OpenRouter for text classification, and Hugging Face for NSFW image scoring and voice transcription. These requests are transient — FreeGuard keeps only its encrypted ledger, not the raw content. Discord itself receives all data you send to it under the platform's own privacy terms. Each provider's policies apply to the brief moment your content passes through their service.
Strikes expire after the configured strikeExpiry; bans lift automatically after banDurationHours unless a staff member extends them; appeal records are retained until the appeal is resolved; the audit log is retained until the operator prunes it. Configuration persists until changed. You can request earlier deletion at any time (see Section 7).
You may request access to, a copy of, or deletion of data FreeGuard holds on you. On a self-hosted instance, contact the server admin who operates it — they have the tooling to export or purge encrypted records. On the public instance, contact the operator through the support server. Deletion requests are honoured within a reasonable timeframe for the records we can attribute to your account.
FreeGuard is designed for servers that comply with Discord's age requirements (13 years and older). The bot does not knowingly collect personal information from children beyond the IDs required for moderation, and no additional account, email or payment data is requested at any point.
If data practices change, this document is updated with a new effective date and the change is called out in the release notes. Material changes to the public instance are announced in the community server before they take effect.
Privacy enquiries about the public instance reach us through the official FreeGuard support community. Matters concerning a self-hosted instance are handled by that instance's operator — check with your server admin first, as they control the data.
Two minutes from invite to live protection. Free forever — no credit card, no trial clock.