One bot, every model. MultiVAC routes your conversation to OpenAI, Gemini or Claude, makes images and video, reaches for real tools when it needs facts, and remembers only what you let it.
// an example exchange
@you is it going to rain in raleigh this weekend? also summarize this link
⟳ get_weather · summarize_url
MultiVAC Saturday stays dry; Sunday has a 60% chance of showers after 2 PM. The article argues …
@you how can entropy be reversed?
MultiVAC INSUFFICIENT DATA FOR MEANINGFUL ANSWER.
Talk to it like anyone in the server. It keeps the thread, follows replies, and picks the model the task needs.
Generate, edit and describe images, analyse attached files, and work from reference images.
Sora and Veo generation, image-to-video from a reference, Sora remixes, and prompts drawn from the message you reply to.
Web and reverse-image search, weather, stocks, URL summaries, YouTube transcripts and repository inspection, inside bounded loops.
Facts you ask it to remember, saved behaviour instructions and optional long-term recall, all kept per user and forgettable on request.
Live progress for long jobs, with real phase details and estimates, and a status record of exactly which tools ran.
/model_defaults; server managers can set a ceiling.Remembered facts, instructions and recall are scoped to the person, channel and server they came from. Traces of what the bot did keep orchestration metadata and public evidence links, never your prompts, attachments, memories or credentials.
Read the privacy policy and terms of service, or write to [email protected].