Who allowed it?
The person the agent acted for, and the agent, even when nobody was at the keyboard.
AgentGate by JUTEQ
Your agents acted last quarter. Can you prove what they did?
AgentGate is an audit trail and decision point for the AI agents your teams built. It records who allowed an action, what the agent did, and under which policy. In private beta. Runs inside your environment.
The demo is a controlled scenario: an agent tries an action outside its task, and you see the decision and the record.
What it answers
Three questions a log file cannot answer.
What did it do?
The action as attempted, and what it touched.
Under which rule?
The policy and version in force at that moment.
In the field
65%
had an incident involving an AI agent in the past twelve months
82%
discovered agents in their environment they did not know about, in the past year
38%
require a human to approve when an agent exceeds its scope
11%
block that action automatically
Cloud Security Alliance, Autonomous but Not Controlled, Jan 2026, n=418
We record what your agents actually did, and decide at the moment of action whether it goes ahead.
Exhibit
The same blocked action, as a log line and as a record.
A / a minimal tool log
2026‑09‑03T02:14:07Z INFO agent_action
agent=agent‑7 status=error
agent=agent‑7 status=error
A tool ran and failed. Nothing about who, what, or why.
B / the same moment, as a record
- recorded
- 2026‑09‑03T02:14:07.339Z
- on behalf of
- a.okafor@bank.example
- agent
- ops‑reconcile/agent‑7, autonomous
- attempted
- export of customer records from production
- decision
- denied · production export, no human present
- policy
- v14, prod‑data‑export
- content
- kept where the agent runs
Illustrative specimen.
Scope
What AgentGate holds.
The record
Shows if it has been altered, and your auditor can verify it without relying on us.
The decision, at the act
Made where the action runs, including for agents working with nobody watching.
Runs inside your environment
The content your agents touch stays where they run. Deploy it in your own environment, or use our hosted service in Canada.
The record is built for the questions auditors and regulators ask. Our first mapping is to Canada's OSFI E-23.
On the calendar
Record-keeping rules are arriving. Canada's OSFI E-23 applies from 1 May 2027, and the EU AI Act's automatic logging requirement from 2 December 2027 for most high-risk systems.
Questions
Questions buyers ask first.
Which agents does it cover?
Agents your own teams built and run in production. Coding agents come next.
Why AgentGate?
AgentGate is designed for the agents your teams built, whichever environment they run in, with a record your auditor can check without trusting us.
What is still being worked out?
The record and the decision point are in place today. We are still establishing, with our beta customers, how far blocking can go without stopping legitimate work, and how consistently one decision holds across cloud, on-premises and hybrid at once. We tell you this up front, and show you what is covered in your environment before you deploy.
Does our data leave our environment?
The content your agents touch stays where they run. The record stays in your environment, or in our hosted service in Canada. You choose.
How long, and what do you need from us?
About six weeks to your first agents governed, with blocking for the action types we agree. You provide an owner, access, and a few hours a week from security, risk and platform.
What does it cost?
A fixed fee, agreed before we start.
What happens when the beta ends?
You keep what you deployed. General availability terms are agreed when you join.
Will you go through our security review?
Yes, before you commit to anything.
Access
Join the private beta
We are deploying with a limited number of organizations, starting with those who have something to prove now.
- A person reads every request and replies within a few days if the beta fits.
- Prefer to see it first? Book a 30-minute demo.
Developers on coding agents too? See the identity rollout.