Building an AI Discord Bot with n8n: Native Node, Slash Commands and Message Trigger

Many teams that use Discord as a community or support space want to plug an AI agent into it, and many hit the same wall: unlike Telegram, n8n has no native trigger that listens to incoming Discord messages. That's not a dead end, but it changes the architecture you need to build. This guide details what the official Discord node can do on its own, how to wire slash commands through a plain n8n Webhook, and how to get a full conversational bot with a community node — without inventing anything: every limitation described here is real and verifiable in n8n's and Discord's official documentation.

What the native Discord node can do (and can't do)

n8n's Discord node (n8n-nodes-base.discord) covers two connection modes:

• Webhook: the simplest mode. A Discord webhook (created from a channel's settings) lets you post messages from n8n without configuring a bot — ideal for one-way notifications (workflow alerts, daily reports).
• Bot: with a bot token, the node handles sending messages and embeds, creating and managing channels, members, and the Send and Wait for a Response operation — the same human-approval mechanism we detail for Slack in our article on the Wait node and interactive buttons, which transposes to Discord to validate a sensitive action directly inside a channel.

What this node does not do: trigger a workflow when a message arrives on a channel. There is no official "Discord Trigger" equivalent to the Telegram Trigger described in our AI Telegram bot guide. That's the key distinction to understand before you start.

Option 1: slash commands through the n8n Webhook (no community node)

Discord exposes an official, documented channel for receiving events without a persistent bot: the Interactions endpoint, used for slash commands (/question, /summary…). This channel works through a plain HTTP call, so it wires directly into a standard n8n Webhook node.

• Create the application on the Discord Developer Portal, register a slash command through the Discord API, and set the n8n Webhook URL as the "Interactions Endpoint URL".
• Verify the signature: every request is signed with Ed25519 (headers X-Signature-Ed25519 and X-Signature-Timestamp). A Code node must validate this signature before any processing — the same reflex detailed in our article on securing n8n webhooks, applied here to a Discord-specific format.
• Reply within 3 seconds: Discord requires an almost-instant acknowledgement. An LLM call almost always exceeds that window. The robust pattern: reply first with a type 5 ("deferred response"), then, once the AI agent has finished, send the final answer via a PATCH call to the Discord API in a separate HTTP Request node — the workflow keeps running in the background without ever blocking the initial acknowledgement.

This approach has a clear advantage: it depends on no third-party package, only native n8n nodes (Webhook, Code, HTTP Request). It fits an assistant driven by explicit commands ("ask the knowledge base a question") well, less so a bot that needs to react to the whole conversation.

Option 2: a full conversational bot with a community trigger node

For a bot that responds to every message (not just a command), you need to listen to the Discord event stream continuously — something the interactions endpoint doesn't cover. The solution is a community trigger node (search "discord trigger" in your self-hosted instance's Community Nodes; several community-maintained packages exist). Once installed:

• Create the bot on the Developer Portal, enable the necessary Privileged Gateway Intents (notably Message Content), and grab the token — store it only in n8n credentials, never in plain text inside a node.
• The trigger node fires the workflow on every message received on watched channels, with the channel and author IDs.
• Wire up an AI Agent node with conversation memory, using the channel ID (or the user ID for direct messages) as the Session ID — exactly the principle detailed in our article on conversation memory for an AI agent in n8n, so each channel keeps its own history. For LLM provider setup, see our guide to connecting Claude or GPT to n8n.
• Reply with the native Discord node (Bot mode, send-message operation) in the original channel.

A community node depends on third-party maintenance: before putting it into production, test it on a dedicated personal Discord server and check its compatibility with your n8n version.

Concrete use cases

• Community support assistant: the bot answers recurring questions from a vectorized knowledge base — exactly the architecture of the chatbot RAG with citations workflow, rewired to Discord instead of WhatsApp or a web widget.
• Report triage: in an active community, moderation reports come in bursts and are hard to trace once the conversation has moved on. A study by Jiang, Kiene, Middler, Brubaker and Fiesler published in Proceedings of the ACM on Human-Computer Interaction (CSCW 2019) on moderation in voice-based Discord communities points out exactly this: gathering evidence of rule violations is one of moderators' main pain points, since live communication leaves few usable traces. A bot that logs every report with a timestamp in Supabase, then notifies the moderation team, addresses this directly — a pattern close to our Compliance & Audit Pack.
• Workflow notifications: a dedicated channel that alerts the technical team when a sensitive n8n workflow fails, with a Send and Wait for a Response button to approve a manual retry.

Discord or Telegram: which one to choose

Criteria	Telegram	Discord
Native n8n trigger	Yes (Telegram Trigger)	No — community node or interactions webhook
Setup	A few minutes via @BotFather	Developer Portal, intents, OAuth2 invite
Natural format	1-to-1 messaging or small groups	Community servers, topic channels
Typical use case	Internal assistant, notifications	Community support, moderation, server bot

For a quick internal assistant, Telegram remains the shortest path. For a bot that lives inside an organized community of channels, Discord is the right ground — at the cost of a slightly more technical setup.

Guardrails before opening the bot

• Systematic signature verification on the interactions endpoint: without it, anyone who knows the webhook URL can trigger the workflow with arbitrary data.
• LLM cost tracking per channel or server, especially for a bot open to an entire community — the method is detailed in our article on tracking AI call costs in n8n.
• Prompt injection: a public channel exposes the agent to messages designed to hijack its instructions, especially if it has tools. Our prompt injection guardrails apply directly here.
• Rate limiting on Discord's side: the API enforces its own per-route limits; a very active bot must handle 429 responses with backoff, on the same principle as our article on handling AI API rate limits.

In summary

An AI Discord bot with n8n is built from three distinct building blocks: the native Discord node to send (plain webhook or full bot), the standard n8n Webhook for slash commands through the signed interactions endpoint, and a community trigger node for a bot that reacts to the whole conversation. Unlike Telegram, none of these is a five-minute shortcut — but the result, an agent with per-channel memory, plugs just as well into a RAG knowledge base. If you'd rather start from a proven foundation than a blank page, the RAG Assistant Pack (€119) provides the document ingestion and the citation-backed RAG chatbot ready to rewire onto Discord.