001
PLATFORM
The control room for every agent you run.
Connect your tools, write the limits, and watch the work happen. Everything an agent does is visible, and nothing risky happens without you.
NIGHT FLOOR · PLATFORM
+
001
HOW AN AGENT WORKS
Four gates between an idea and an action.
The same path every time, for every agent, so there is nothing to guess.
Gate 01
Agent proposes
It writes down the action it wants to take and the reason, using only the tools you connected.
Gate 02
Rules check
Your limits are applied. Anything outside them stops here and is sent to a person.
Gate 03
You approve
Big or unusual actions wait for a yes. Routine ones you have allowed go straight through.
Gate 04
Done and logged
The action runs and the result is written to the audit log with the time and the reason.
(02)
Rules
Limits you can read out loud.
A rule is one sentence with a number in it. Write it, test it against last week, then switch it on.
New rule
Draft
You typed
Refunds under $100 are fine, anything bigger asks me.
NextAI turned it into
IF
refund amount is under $100
THEN
approve
ELSE
refund amount is $100 or more
THEN
ask a person
Tested on last week: 214 runs, 6 would change.
Audit log / Support agent
Today
07:12:03
Opened ticket 9902
READ
07:12:06
Checked carrier status
READ
07:12:08
Sent tracking message
WROTE
07:19:41
Closed ticket 9902
WROTE
07:22:10
Refund $212 on order 20417
HELD
(03)
Audit log
A record you can hand to an auditor.
Every read, decision and change is written down with its time and its reason. Filter it, search it, export it.
(04)
Approvals
A yes or no from anywhere.
Approval cards show the same facts as the dashboard. Approve, skip or ask a question from your desk or your phone.
Sales agent
Waiting
Send quote to Halden Works
$18,400, 12% under list
Approve
Skip
Support agent
Waiting
Refund order 20417
$212, above the $100 limit
Approve
Skip
Billing agent
Waiting
Pay invoice #4790
Looks like a duplicate of #4712
Approve
Skip
005
CONNECT
Twelve tools connected today, more every month.
Sign in once. Choose what each agent can read and what it can change.
→
NEXT STEP
See it running on your own tools.
Book a 30-minute demo. We connect a sandbox and show one agent doing one of your real tasks.