Your AI command center
Chatting is cheap. When you ask for real work, Sirius spins up a task below. Say "every day at 6pm…" and it becomes an Automation. Images, audio and video links render inline.
You talk to Sirius — the orchestrator. Sirius picks the right specialist for each job, hands it off, and reports back. The specialists also feed each other: research finds the angle, content turns it into posts, and every plan they make drops into Action Items so you and the agents can see what's next.
Agents can only queue — they have no path that sends. Review the copy, then Approve to publish via Publer (29 connected accounts) or GoHighLevel.
Or just ask in chat — "give me a sales recap every day at 6pm" — and it shows up here. Automations run on the metered API lane by default so they can fire unattended.
Saved in this browser and handed to the send engine once one is connected. Setting them now means outreach starts inside your limits rather than being tuned after the first send.
Sequences, versions and per-campaign quotas. Populated by the send engine.
Imported prospects, segments and enrichment columns.
Inbound answers, sentiment and anything needing you personally.
SPF, DKIM, DMARC, bounce rate and blacklist checks per sending domain.
Read live from the agent history log (orchestrator/history/). Outreach view filters to email, lead and campaign activity; switch to All activity to see everything Sirius ran in the same window.
orchestrator/projects.json.
Runs always-on as com.sirius.voice. Also on Telegram once the bot token is in ~/.env. Cost: about $0.10 to $0.15 of ElevenLabs per minute Sirius speaks; listening is free. When you are done talking, switch to the Chat tab for Office mode.
library tools.
Everything exchanged between you and the agents — your uploads and their finished work.
—
No character picked yet — Sirius is running in its default voice.
Facts and preferences Sirius picks up from your conversations. They shape every reply and task. Delete anything it got wrong.
Connect your accounts so agents can work them. Values are write-only — saved on this machine, never shown back, never leave it. Use them in MCP servers as ${cred:service.key}.
Every capability your agents can be granted. Add more by dropping a *.tool.js in orchestrator/tools/ or adding a server to mcp-servers.json, then hit Rescan. Grant them to an agent in the Agents tab.