Concepts — the six things a user thinks about
The whole product is built on six ideas, each in plain language. A non-technical person should understand all six in a couple of minutes. Nothing else is a user concept; everything else is a detail behind one of these.
Each concept below has: a one-line definition, an everyday analogy, what the user sees, and what the user does.
1. Agents
What it is: the AI agents and automations you run — the things doing work on your behalf.
Analogy: like employees or apps you've hired to do a job — each has a name, an owner, and a set of things it's trusted to do.
What you see: a list of your agents, each with an owner, a risk level (Low / Medium / High), a status (green/amber/red), and what it's spent this month.
What you do: add an agent, open one to see what it's up to, or stop one.
A new agent starts able to do nothing. You decide what it can do. That's on purpose — see Permissions.
2. Permissions
What it is: what each agent is allowed to do. Everything is off until you turn it on.
Analogy: like app permissions on your phone — an app can't use your camera or location until you say yes. Same here: nothing is allowed until you allow it.
What you see: a plain list of capabilities for the agent — for example "Read customer data," "Send emails," "Issue refunds" — each switched Off by default. When you turn one On, you see a short note about how risky it is and whether it will need your approval each time.
What you do: turn capabilities on one at a time, only the ones the agent actually needs.
Safe by default. An agent with everything off cannot do anything. That's the safest possible state, and the product tells you so.
3. Budgets
What it is: how much an agent can spend, per month.
Analogy: like a prepaid card or a monthly allowance — you set the limit, you can watch it fill up, and you decide what happens when it runs out.
What you see: a spend cap, a usage bar showing how much of it is used, and your choice of what happens when the cap is reached.
What you do: set the monthly cap, choose the behavior when it's hit — just alert me, slow it down, or stop it. (A manual reset is coming.)
Budgets are a smart limit, not a to-the-cent wall. Spend is watched continuously and acted on fast, but a small overshoot is possible before a hard stop kicks in. Set the cap as "stop around here."
4. Approvals
What it is: risky actions an agent pauses on, waiting for a human's one-click yes or no.
Analogy: like a manager's sign-off — before someone does something that can't easily be undone, they check with you first.
What you see: an inbox of simple cards: "[Agent] wants to [do this] ([why it needs you])." Each shows who requested it and when, with Approve and Deny buttons.
What you do: approve or deny. The agent waits until you decide; if you don't decide in time, it expires safely (no action taken).
The riskiest capabilities are set to ask for your approval every time, automatically. Choosing which capabilities need approval yourself is coming.
5. Stop
What it is: an instant kill switch for an agent.
Analogy: like the big red emergency-stop button on a machine — one press and it halts, completely.
What you see: a clear Stop button on every agent, and a plain confirmation: "Stop [agent] now?" After you confirm: "Stopped. It can't do anything else."
What you do: press Stop, confirm, done. The agent is fully stopped — its work halts, its access is pulled, and it's blocked from acting again. (Restarting a stopped agent is coming — today, stopping is one-way.)
Stop is thorough and instant. It doesn't just pause one task — it holds back the whole agent across everything it could do.
6. Proof
What it is: a board- and auditor-ready report showing that everything was governed — and that the record hasn't been tampered with.
Analogy: like a receipt with a tamper-proof seal — it shows exactly what happened, and the seal proves no one quietly changed it afterward.
What you see: a Verified badge and a plain summary — "X actions, all governed; Y approvals; nothing tampered with." Plus an Export button for a file you can hand to a board, an auditor, or a regulator.
What you do: glance at the badge for reassurance, or export the full report when someone needs evidence.
Proof shows what was governed and that the record is complete and unaltered. It's the evidence behind the calm "everything is governed" message you see on the home screen.
How the six fit together (one paragraph)
You run Agents. Each one can do nothing until you grant Permissions, and can spend only up to its Budget. The riskiest actions stop and wait for your Approval. If anything looks wrong, you Stop the agent in one click. And the whole time, Proof is quietly recording everything — verified, sealed, and ready to show anyone who asks. That's the entire product.