Discord: built and not yet live, with two honest caveats

Short version: Allinaa has built a Discord connection and it is not yet live. Nothing in the way of it is Allinaa's approval - a business supplies its own application and bot - but there are two caveats worth stating before anybody plans around it, and they are policy questions rather than code.

Status, and what the two caveats are

Discord, which is not yet live, needs no Allinaa platform credential: the business creates its own application in the developer portal and supplies the bot token and public key. The code exists and is tested against a local stand-in for the platform's gateway and REST interface. What has not been established, and is stated here rather than glossed over, is whether the platform's own policies allow a third-party service to hold a business's bot token and to process community message text with AI. Until somebody has read those terms for your situation, treat this as built rather than recommended.

What the built connection does

• Direct messages to the bot are answered like any other conversation, within your settings.
• Mentions in a server are treated as public: they are stored as a public message and any reply waits for a person to approve it, because a reply in a server is visible to everyone in it.
• Product cards are sent as embeds, where the name, price, availability and link come from your own catalogue.
• Attachments are fetched from the platform's own content network and follow the same file rules as the website chat - type sniffed from the content, size and quota limits, malware checks, private storage, retention. See files customers send.
• Interaction signatures are verified cryptographically, so a request that did not come from the platform is refused.

What it deliberately does not read

It does not request the privileged permission that would let it read every message in a server. That means it sees direct messages and messages that mention the bot, and nothing else - not the general conversation of your community. That is a deliberate choice rather than a limitation to be lifted later: a service that reads a whole community's chat is a different product with different obligations, and most communities would not want it.

Why a community is a harder audience than a website

This is the part worth thinking about before the technical part. A website visitor is a stranger with a question. A community member is somebody who will see every answer the bot gives to everybody else, will test it, and will screenshot a bad one. Three consequences:

• Public replies need approval, which is how the connection is built - no automatic posting into a server.
• Repetition reads badly. The same canned answer three times in a visible channel looks worse than silence.
• Your boundaries will be probed on purpose. A community will find the question the assistant answers badly, for sport. Write the refusals before you launch, not after. See writing the way your chatbot should speak.

One connection, one process, and what that means

The platform's gateway needs a persistent websocket, which is a different shape from every other connection on this site. The connection resumes its session after a restart so the platform replays anything missed, closes cleanly on shutdown, backfills recent direct-message threads after a fresh session, and respects the platform's daily limit on starting sessions. One honest consequence: because a stop for a permissions error is remembered in memory rather than on disk, a restart can produce one more reconnect attempt and one more notification before it settles.

What to use today

If your reason for wanting this is that your customers are in a community rather than on your website, Telegram is live-verified and works the same way for direct messages - a business connects its own bot, and the assistant answers within the settings you choose. The website chat is the other live channel. See Telegram, the website chat and the channel overview.

Questions people ask about this

Can the assistant moderate my server?

No. It is not a moderation tool, it does not read the general conversation, and it has no moderation actions.

Could it answer in a support channel automatically?

Not as built: anything visible to a server waits for a person to approve it. That is the right default for a public room, and it is what stops one wrong answer being read by five hundred people.

Will it ever be live?

Honestly, that depends on the policy question above and on whether businesses ask for it. There is no date, and inventing one would be the kind of claim the rest of this site avoids.