AntiLink Guard OSS

Open-source Discord anti-phishing and link moderation framework.

A real detection and policy engine for Discord link moderation - self-hosted, open-source, and yours to run, read, and modify.

Quick start · Features · Architecture · Docs · Roadmap

---

The problem

Discord communities get hit with the same handful of attacks constantly: phishing links dressed up as free-Nitro giveaways, scam Discord invites, punycode/homoglyph domains that look identical to legitimate ones at a glance, and links deliberately obfuscated (hxxps://, example[.]com, zero-width characters) to slip past naive keyword filters. Most self-hostable anti-link bots do a plain .includes("http") check and call it done.

	Naive keyword bot	AntiLink Guard OSS
Detection	Substring match on http	Real link extraction: markdown, bare www., Discord invites
Obfuscation	Bypassed by hxxps://, [.], zero-width chars	De-obfuscated before scoring
Lookalike domains	Not detected	Punycode + Latin/Cyrillic/Greek homoglyph checks
Decision	Binary: delete or ignore	Scored verdict (ALLOW/WARN/BLOCK/QUARANTINE)
Enforcement	Fixed	Configurable ladder: log → warn → delete → timeout
Failure mode	Crashes or silently no-ops on missing permissions	Checks permissions first, fails safe, logs why
Message content	Often logged in full	Audit log type has no field that can hold it
Extensibility	Fork the script	Typed packages: swap storage, embed the engine, script the CLI

Features

Everything below is implemented and tested in this repository:

• 🔗 Real link detection - not just https?://. Catches markdown links, bare www. domains, Discord invites, and links deliberately obfuscated with hxxp://, example[.]com, or zero-width characters.
• 🕵️ Classification, not just matching - domain allow/blocklists, a guild-suppliable known-phishing list, common URL shorteners, punycode hostnames, and Latin/Cyrillic/Greek homoglyph mixing, plus your own custom regex rules.
• ⚖️ A real policy engine - every message gets a score and one of four verdicts (ALLOW/WARN/BLOCK/QUARANTINE), mapped to an action through a log → warn → delete → timeout enforcement ladder you control per server.
• 🔐 Permission-safe by construction - the bot checks its own Discord permissions before every delete/timeout and fails safe (logs and skips) rather than crashing or assuming success.
• 🗄️ Your database, your choice - memory (for tests), SQLite (the self-hosting default), MySQL, or PostgreSQL, all behind one interface.
• 📊 Metadata-only audit logging - the audit log type has no field capable of holding message content, by design. See docs/privacy.md.
• 🧰 A real CLI - antilink scan/test-url run the exact same engine the bot uses, entirely offline, no Discord connection required.
• 🖥️ A minimal local dashboard - read-only, unauthenticated by design, meant for localhost (see apps/dashboard-lite).

No hardcoded "known phishing domain" database is bundled anywhere in this codebase - that data is always something you (or a feed you choose) supply. See docs/rules-engine.md.

Architecture

flowchart TD
    subgraph Discord
        MSG[messageCreate event]
        CMD[Slash command interaction]
    end

    subgraph "@antilink-guard/discord-bot"
        MSG --> PIPE[Moderation pipeline]
        CMD --> ROUTE[Command router]
        PIPE --> ENFORCE[enforce.ts<br/>permission-gated delete/timeout]
        ENFORCE --> MODLOG[Mod-log embed]
    end

    subgraph "@antilink-guard/core"
        EXTRACT[extractLinks<br/>de-obfuscation]
        CLASSIFY[classifyLink<br/>scoring]
        POLICY[evaluateMessage<br/>verdict + action]
        EXTRACT --> CLASSIFY --> POLICY
    end

    PIPE --> EXTRACT
    ROUTE --> POLICY

    subgraph "@antilink-guard/storage"
        DB[(SQLite / MySQL / PostgreSQL)]
    end

    POLICY <--> DB
    ENFORCE --> DB

    CLI["@antilink-guard/cli<br/>(scan / test-url / doctor)"] --> EXTRACT
    DASH["apps/dashboard-lite<br/>(local, read-only)"] --> DB

Loading

packages/core has no dependency on Discord or on any storage backend - it's a pure detection/policy library you can use standalone (the CLI does exactly that). See docs/rules-engine.md for the full pipeline and docs/api-reference.md for every package's exports.

Quick start

git clone https://github.com/timeout187/AntiLink-Guard-OSS.git
cd AntiLink-Guard-OSS
pnpm install
pnpm run build

cp apps/example-bot/.env.example apps/example-bot/.env
# edit apps/example-bot/.env: DISCORD_TOKEN, DISCORD_CLIENT_ID

pnpm --filter @antilink-guard/example-bot run register-commands
pnpm --filter @antilink-guard/example-bot run start

Full walkthrough (Discord Developer Portal setup, required permissions, Message Content Intent): docs/getting-started.md.

Docker Compose

cp apps/example-bot/.env.example apps/example-bot/.env
docker compose up --build -d

Builds the whole workspace inside the container and runs the bot under a non-root user, with a persistent volume for its SQLite database. See docs/self-hosting.md for switching to MySQL or PostgreSQL instead.

Usage

/antilink enable
/antilink mode block

/allowlist add domain:cdn.example.com
/blocklist add domain:free-nitro-gift.ru

/invites block-all enabled:true
/invites allow invite:https://discord.gg/your-partner-server

/logs set-channel channel:#mod-log

/testlink url:https://xn--e1aybc.xn--p1ai
/config export

Every admin command requires Manage Server; /testlink is open to everyone. Full reference: docs/configuration.md.

Prefer the command line? The exact same engine, no Discord required:

antilink scan "check out hxxps://free-nitro[.]ru/claim"
antilink test-url https://bit.ly/abc123
antilink doctor

Privacy & security

• No telemetry. This project doesn't phone home anywhere.
• Metadata-only audit logs - see docs/privacy.md for exactly what's stored (and what deliberately isn't).
• Threat model: docs/threat-model.md - what this framework does and doesn't defend against.
• Reporting a vulnerability: see SECURITY.md - please don't open a public issue for security problems.

Repository layout

apps/
  example-bot/      a working self-hosted bot built on the packages below
  dashboard-lite/    a local, read-only dashboard
packages/
  core/              URL/invite extraction, classification, policy engine
  storage/           memory / SQLite / MySQL / PostgreSQL adapters
  discord-bot/       the discord.js v14 adapter + moderation pipeline
  cli/               the `antilink` command-line tool
docs/                the guides linked throughout this README

Contributing

Contributions are welcome - see CONTRIBUTING.md for the development setup (pnpm workspace, build order, testing) and CODE_OF_CONDUCT.md. Governance model: GOVERNANCE.md.

Roadmap

See ROADMAP.md for what's shipped, what's next, and what's explicitly out of scope for this project.

License

MIT.

Background

AntiLink Guard OSS is based on lessons learned from running AntiLink, a Discord moderation bot used across 100+ servers. This repository extracts the reusable moderation, URL detection, and policy engine into a self-hostable open-source toolkit for the community.

This repository contains only the open-source framework described above. It is not affiliated with, and does not include, any hosted or commercial product that happens to share part of its name.