Getting started.
Invite the bot, send your first message, and lock him to specific channels or roles.
Invite the bot
Open the invite, choose a server you own or manage, and approve the default permission set. No signup, no dashboard step, no credit card.
- Open the invite link in a browser where you're logged into Discord.
- Pick a server - only servers where you have Manage Server permission show up.
- Authorize the default permission set. You can tighten or loosen these later in Discord's role settings.
Read/send messages, manage messages (for purge and automod), connect/speak in voice (for /voice and /play), embed links, and attach files. Nothing destructive without explicit slash command + Discord permission gate.
Send your first message
There's no setup wizard. Just say hi.
- Mention him -
@Jarvis what's the weather like in chicago - Slash command -
/jarvis prompt: explain monads in two sentences - Wake word - type or say
jarvis(orgarmin) and he responds globally where he's allowed - Reply chain - reply to one of his messages and the thread continues
Jarvis only reads channels where someone explicitly talks to him. He does not lurk surrounding conversation. If you reply to someone else's message while talking to Jarvis, he reads that single replied-to message for context - nothing else from the channel.
Restrict where he listens
By default Jarvis responds anywhere he can read. Most servers want him in one or two channels. Two commands cover it:
/channel set channel: #ai-chat
/channel clear
/role add role: @AI-Users
/role remove role: @AI-Users
Channel lock: after /channel set, Jarvis only replies in that channel. Use /channel clear to unlock.
Role gate: after /role add, only members with the listed roles trigger Jarvis. Useful for member-only AI access.
Add him to voice (optional)
Join a voice channel, then run /voice. Jarvis joins, listens, and replies out loud when the wake word fires. Run /leave to disconnect him.
Voice runs on NVIDIA NIM - Parakeet for speech-to-text, a NIM TTS model for the voice. Raw audio is never recorded; only post-wake-word transcription crosses the wire.
/play is for music, not voice chat. Only /voice puts Jarvis into listening mode. Two different systems sharing the same audio channel.
Where to next
- Commands reference - what every category does and when to reach for it.
- Integrations - AI providers, voice stack, music sources.
- All commands - the auto-generated full list.
- FAQ - pricing, privacy, model choice, removal.
Drop a question in the AGIS support channel. Most things get answered within the day - the maintainer reads every message.