Trellvix

The questions worth asking first.

This is the part a serious buyer actually reads: what to hand over first, how much setup it takes, where the control boundary sits, and whether the product is trustworthy enough for real operational work.

Questions

The business-buyer questions worth asking first.

This is the part a serious buyer actually reads: what to hand over first, how much setup it takes, where the control boundary sits, and whether the product is trustworthy enough for real operational work.

MemoryDo I have to re-explain context?
No. It keeps what you tell it and connects it up, so “my accountant” and “Priya at Deloitte” are one person rather than two facts. It also keeps where each thing came from - whether you said it, it read it, or it worked it out - and it will tell you which.
PositioningWhy not just use Claude or ChatGPT?
Those tools are useful components. But turning them into something that behaves like a reliable chief of staff usually means prompts, connectors, setup, and a lot of manual supervision. Trellvix is built around the job itself: operational memory, prepared work, follow-through, approval, and traceability.
ControlCan it act without asking?
No. Anything that spends, sends, or can’t be undone waits for your explicit approval. Trellvix proposes; you decide.
PrivacyHow do you handle sensitive business context?
Only if the boundaries are clear. Your agent is isolated to you, opened by your passkey, and approval-gated on anything irreversible. And if there is something you do not want us to read at all, you can seal it so we hold the bytes and cannot open them.
PricingIs it free to start?
Yes - free to start, no invite and no waitlist. Fair-use limits keep it sustainable while it’s early; nothing you have to think about in ordinary use.
OwnershipWhat if I want to leave?
Export everything it knows in a single move and go. If you want it gone as well, closing your agent deletes all of it - what it learned, the record of what left, anything you sealed, and the passkey itself - and retires the name so it is never given to anybody else. Take your copy first, because afterwards there is nothing left to export. No lock-in, ever - that’s the point.

Guides

How three kinds of people run one agent.

One agent, one routine, three ways of setting it. Each guide is written from what the agent can actually do today, and each routine line below is one it will read as written.

FounderYou are the bottleneck, and the inbox knows it.

Tell it who matters first: the investors, the two customers you would take a call from at midnight, your accountant. It keeps each as one person, not a scatter of facts. Then give it one line to run unasked:

every hour - look for anything that needs you now

Anything that spends or sends waits for you, with the work already done, so an approval is a glance rather than a task. Open Approvals when you have a minute, not when it pings.

OperatorYou run the desk, and the desk has to run without you.

Give it the standing shape of your week, in words, and check the pane says back what you meant:

every morning - go over what is waiting
every week - look for work that has stalled

Everything that leaves goes through one door and lands on the record, so when someone asks what went out and why, the trace already has it. Seal what it should hold and never read; export the rest whenever you want a copy.

Solo professionalYou are the whole company, and context is the job.

Talk to it the way you would to a colleague who never forgets: the client, the deadline, the thing you promised on a call. Ask it later and it says where each fact came from, whether you said it, it read it, or it worked it out. A light routine is enough:

every evening - look for anything that needs you now

Nothing goes out without you. Nothing is taken on faith. And if you leave, everything it knows leaves with you.

Meet the chief of staff your company never had.

Read the twelve statements