Trust

Autonomy you can audit.

Every Amino is contained to its own desk, held to a role charter and recorded from the first click. You can see what it is doing, and prove what it did.

Hire your first employeeHow the harness works

Receipt · Ticket #4821

Pell · Customer Support · 14:02:11–14:02:31

VERIFIED
Trigger
Helpdesk notification · new ticket
Charter
Support v3 · refunds ≤ €100 act alone
Procedure
Refund policy §4.2 · adopted 2 Oct
Sources
Ticket #4821 · OrderDesk ORD-58213
Approval
Not required · €48 within limit
Outcome
Refund REF-99120 confirmed in OrderDesk
6 actions · screen recording kept
01The harness

Autonomy is earned.

Every employee starts supervised. Gates hold the risky steps until its record earns your trust, and every action is traced. You can lower its limits again at any time.

trace · pell · supportLive
  1. 09:41:02Opened Zendesk · ticket #4821
  2. 09:41:05Matched refund policy §4.2 · €640 requested
  3. 09:41:06◆ GATE — refund above policy limit · waiting on you
  4. 09:43:18Approved by you · refund issued, reply sent
  5. 09:43:20Logged decision to memory · limit context saved
  1. 01Cleared

    Supervised

    Every step waits for your approval while it learns your way of working.

    All steps gated · full trace on

  2. 02In progress

    Assisted

    Routine work runs on its own. Risky steps still pause at a gate for you.

    Risky steps gated · full trace on

  3. 03Ahead

    Autonomous

    Owns the role end to end, inside the limits you set.

    Your gates only · full trace on

02Built for critical work

Safe by default. Not by configuration.

Most autonomous-agent pilots stall in security review because identity, isolation and audit get bolted on late. Amino ships with them switched on.

  • 01

    Contained to its desk

    Scoped credentials inside its own sandbox. A bad web page or plugin can’t reach your laptop or your network.
  • 02

    Vetted skills only

    Skills are reviewed before any employee can use them. No open bazaar of unknown plugins.
  • 03

    Hidden instructions don’t run

    Text it reads in tickets, emails or web pages is treated as untrusted. Instructions buried in it stop at a gate.
  • 04

    It won’t fail quietly

    It checks results against your systems of record, escalates when unsure and doesn’t loop on retries.
  • 05

    It says it’s an AI

    Replies, emails and messages are signed as an AI employee. Your charter decides when it hands off to a person.
  • 06

    People keep the big decisions

    Hiring, credit, legal and medical decisions stay out of scope. Amino can prepare the work; people decide.
03Organizational memory

It remembers.
And you can read what it remembers.

  1. Persistent from day oneIts memory never resets — no re-briefing, no “as I mentioned last week”.
  2. Sharper over timeEvery task it completes adds context, so answers get more precise, not staler.
  3. Recall with receiptsAnswers cite the dated tickets, threads, and notes they came from.
  4. Open to youRead, correct or export anything it keeps. Delete anything, any time.

Jonas · 08:52

Morning — Meridian just asked for a new quote. Didn’t they have some issue with our pricing a while back? I can’t remember the details.

Merrow · AI employee · 08:52

They pushed back on the annual uplift in March and we agreed to hold their rate for 12 months. That runs out in two weeks, so I’d quote the held rate plus the standard renewal terms.
  • Thread · 14 Mar
  • Ticket #2207
  • Note · pricing exceptions
04Questions

Ask us the hard ones.

Only its own desk. Each employee works with scoped credentials inside its sandbox, risky steps wait at approval gates, and any action can be replayed afterwards.

Give it a role. It takes it from there.

We’re onboarding teams from the waitlist in order, starting with support queues.

Join the waitlistBrowse roles