Stay compliant.
Give every AI agent an owner, a limit and a record. Use one policy engine for people and agents to decide which actions are allowed, need approval or are denied.
THE GAP
Agents are already acting.
Accountability hasn’t caught up.
Access is not governance. As agents work across your systems, there is no single place to set the rules, approve actions or track who authorized them.
Rules in pieces
Each system enforces its own controls. Rules set in one system don’t carry into the next step of an agent’s workflow.
No checkpoint
Actions that need human judgment go ahead with no one asked to approve them.
Accountability after the fact
When something goes wrong, teams reconstruct what the agent did, whose authority it used and who approved it.
Rules in pieces
Each system enforces its own controls. Rules set in one system don’t carry into the next step of an agent’s workflow.
No checkpoint
Actions that need human judgment go ahead with no one asked to approve them.
Accountability after the fact
When something goes wrong, teams reconstruct what the agent did, whose authority it used and who approved it.
Rules in pieces
Each system enforces its own controls. Rules set in one system don’t carry into the next step of an agent’s workflow.
No checkpoint
Actions that need human judgment go ahead with no one asked to approve them.
Accountability after the fact
When something goes wrong, teams reconstruct what the agent did, whose authority it used and who approved it.
AGENT GOVERNANCE
One path from agent to tool.
Every step is governed.
Set the rules, approve actions and track who authorized them, all from one place.
DiscoverDiscover
Find each agent with its owner, connections and tools.
SecureSecure
Issue short-lived, scoped tokens for each task.
AuthorizeAuthorize
Check each call against policy. Tools an agent can’t use stay hidden from it.
ApproveApprove
Send risky calls to a named approver in the NewCore App.
RecordRecord
Log each call against the accountable owner.
RevokeRevoke
Cut off a single session or the entire agent.
DiscoverDiscover
Find each agent with its owner, connections and tools.
SecureSecure
Issue short-lived, scoped tokens for each task.
AuthorizeAuthorize
Check each call against policy. Tools an agent can’t use stay hidden from it.
ApproveApprove
Send risky calls to a named approver in the NewCore App.
RecordRecord
Log each call against the accountable owner.
RevokeRevoke
Cut off a single session or the entire agent.
DiscoverDiscover
Find each agent with its owner, connections and tools.
SecureSecure
Issue short-lived, scoped tokens for each task.
AuthorizeAuthorize
Check each call against policy. Tools an agent can’t use stay hidden from it.
ApproveApprove
Send risky calls to a named approver in the NewCore App.
RecordRecord
Log each call against the accountable owner.
RevokeRevoke
Cut off a single session or the entire agent.
RUNTIME CONTROL
One policy engine for people and agents.
Enforced at runtime.
Set the rules once and apply them to every agent. Routine calls go through without waiting on a person.
Write policy in a sentence
Describe the use case in plain language. Confirm which agents the policy applies to.
Test before it goes live
Replay the policy against recent real sessions. See what would have changed before you save it.
Allow, ask or deny
Let routine calls through, send sensitive ones for human approval and stop anything outside policy.
HUMAN IN THE LOOP
Risky actions wait.
A named person decides.
Keep routine work moving and bring in a person when policy requires it, with the context to decide in the NewCore App.
Monday, September 21
9:41
- 01
Policy flags the action
High-risk actions go to the agent’s owner, a manager or a designated person for approval.
- 02
A person decides
The approver sees which agent is asking, who owns it and what it wants to do, then allows or denies the request. An agent can never approve its own action.
- 03
Access is scoped
Once the approver confirms with device biometrics, access is short-lived and task-scoped. If the request is denied or expires, nothing runs.
- 04
One click revokes
The owner or an admin can cut the agent off at any time.
EVIDENCE, NOT LOGS
Every call on record.
Every record tied to an owner.
Tool calls, sessions and approvals are captured in one audit trail. Evidence is ready before the compliance review starts.
Verifiable
Prove who or what acted, under whose authority and against which resource.
Attributed
Trace each automated call back to the person accountable for it.
Signal
Treat denied access attempts as a primary threat signal, not background noise.
Verifiable
Prove who or what acted, under whose authority and against which resource.
Attributed
Trace each automated call back to the person accountable for it.
Signal
Treat denied access attempts as a primary threat signal, not background noise.
Verifiable
Prove who or what acted, under whose authority and against which resource.
Attributed
Trace each automated call back to the person accountable for it.
Signal
Treat denied access attempts as a primary threat signal, not background noise.