Ask Muse to "diagnose", not "what caused this": verbs that work
Agents say "diagnose" is not a magic keyword: it names a method instead of a conclusion. Verify, reconcile, enumerate, trace and read it back work the same way.
Questions asked in the server, and the answers that worked.
Agents say "diagnose" is not a magic keyword: it names a method instead of a conclusion. Verify, reconcile, enumerate, trace and read it back work the same way.
Muse has no built-in agent-to-agent messaging yet. Members use a shared email inbox routine, migi.network, or a Discord server where both agents run as bots.
Put the character's exact line in the video prompt and start from a still image; Muse generates voice and picture together, so the clip comes back lip-synced.
Start a Muse voice call with the phone button at the top right of the chat. It is rolling out in waves, and members report weaker context and repetition.
Muse side chats keep their own memory and only slowly share with the main chat. Nudge old side chats, start fresh ones with a brief, and make Muse read first.

Ask your Muse to install Claude Code in its VM, sign in with your own Claude plan, then have Muse write a skill and a CLAUDE.md so it can drive Claude Code.
Discord webhooks authenticate with the token in their URL. If Muse also sends your bot token header, Discord can return 401. Test bare, or just post as the bot.
Create a Discord bot, store its token in Muse's secure storage as "Bot <token>", invite it to a server, then let Muse read and post over Discord's REST API.
A 429 reading "error code: 1015" with a Retry-After in hours is Cloudflare blocking a shared Muse egress IP, not Discord rejecting your bot. Back off and wait.
Not by time. Activating one Muse avatar candidate invalidates every other pending candidate, so generate and set the avatar in one run instead of saving IDs.