Assistants wait to be asked. Operators notice.
Every request becomes a ticket with a phase: needs-you, blocked, done, fyi. You see the same queue in WhatsApp, in the logs and in Notion. An agent that acts without a paper trail is unusable in a real company.
Follow-up that actually happens
It sees which deals went quiet, drafts the next touch in your voice, and brings it to you before it sends.
> who has gone quiet this week
6 deals, drafts ready, waiting on your yes
Pipeline hygiene
Call ends, voice note in, deal stage moved and the note logged. No app switching, no end of week cleanup.
> call with Tariq done, moved to proposal
HubSpot deal updated in 4 seconds
The morning brief
A 90 second voice brief before you open a laptop. Pipeline, calendar, and anything that came in overnight.
> brief me
8 deal updates, 3 flagged threads, 2 calls prepped
Research before the call
Company profile, recent news, tech stack, hiring signals. Ask by name, get a structured brief back.
> research Acme before my 3pm
one page brief with signals and talking points
Work with no owner
The handoff nobody picked up. It finds the task, names who it belongs to, and chases it until it closes.
> what is stuck and on whom
4 items, owners named, nudges queued
Coding agents you approve by voice
Spin up a coding agent, watch it work, approve or redirect it from your phone. Nothing ships without you.
> ship the auth fix and show me
branch pushed, diff waiting for your approval
Most chat tools forget everything the moment the session ends
Vega runs a knowledge graph that remembers every person, deal, and decision you have discussed. Say 'prep me for Tariq' and it pulls every interaction, deal note, and research brief tied to Tariq without you re-explaining anything. That is the difference between a chatbot and an ops layer that compounds.
- Person and org graph
- Every contact and company you mention gets a persistent node in the graph.
- Decision memory
- Pricing calls, product commitments and follow-up promises are indexed and retrievable.
- Cross-session continuity
- Conversations do not reset. It picks up exactly where you left off.
Run it yourself, right here
This is the same research engine that briefs you before a call. Type a target and watch it work.
Type a niche + location, hit Find leads.
Try: dentists in Austin · B2B SaaS founders in NYC
Demo only — your run pulls real leads via Apify. Book a call to set up your scraper.
Set up real leadsThe automations I run from my own phone
These are live in my own operation today. The ones marked coming are in build. Every one is driven from a message or a tap, with no laptop in the loop.
Talk to Termius
Pick any live coding session, read its last summary, then voice or type the next instruction. It lands straight in the terminal.
WhatsApp triage and notify
When a background agent needs a decision, it pings you on WhatsApp with the ask, so you unblock it from anywhere.
Apple Shortcuts pack
Tap-to-install signed shortcuts that hit your private webhooks: ask your memory, save a reel, pull approvals, run a doctor.
Okdone fleet control
One decision queue across every running agent. Approve, redirect, or hear the result by voice. Self-hosted, no vendor sees your telemetry.
Inspiration feasibility agent
Drop any reel or idea. It checks feasibility against your stack and waits for a thumbs up on WhatsApp before anything ships.
Voice-note reports
A sub-30-second WhatsApp voice note after each unit of work, with a go or no-go gate before anything irreversible sends.
WhatsApp reply loop
Reply to a notification with a message or a voice note and it continues that exact session. No app switching.
Pocket keyboard
A Wispr-style keyboard that dictates into any iOS app and transcribes on your own server, not in someone else's cloud.
Before you book the audit
The questions people ask before they run this on their own number and their own server.
Is Vega really open source?
Yes. The orchestration layer, skills, memory graph, and wiring are open. The only thing you pay for separately is your AI subscription, which goes straight to the model provider. Everything runs on your own server.
Does this work on my personal WhatsApp number?
Yes. The agent runs on your existing number through a bridge on your own server, so your contacts see messages from your normal number. You do not need a second number unless you want one for the agent.
Do I get locked in to your infrastructure?
No. The skills, the memory layer, and the wiring are yours. If you want to migrate or self-host after setup, I hand you everything. Managed hosting is optional, for teams who do not want to run their own server.
Can my team use it, not just me?
Your team talks to it in the channel they already use. It knows who asked, and it brings decisions to you. It is an operator working alongside your people, not a manager of them.
What does it do without asking me first?
Reading, research, drafting and internal updates run on their own. Anything that leaves the building, an email, a message to a client, a deploy, stops at an approval and waits for your yes. You set where that line sits during the audit.
Ready to see the firm number?
One audit, then a scoped build priced to your stack. No monthly plan, no retainer you cannot leave.