Solutions

Interrupt the paths that matter.

One control loop, applied where fragmented context does the most damage. We start with identity, and the same loop is designed to extend across controls and to AI agents.

Circuit is in development. Scenarios are illustrative; lab figures come from our lab, with fictional data.

First focus

Identity & NHI exposure interruption

Identities — human and non-human — connect everything: cloud, SaaS, workloads, data and AI agents. When a compromised or over-privileged identity lines up with reachability, an exploitable exposure and critical data, Circuit recommends a bounded, preferably reversible change that breaks the path, and proves that it did.

Designed interventions

  • Revoke sessions
  • Remove entitlements
  • Reduce privilege
  • Disable credentials

Why identity first

Identity connects naturally to every other domain and offers practical, often reversible, enforcement actions.

Example scenario
  1. Employee account
  2. Finance agent
  3. Service identity
  4. Production DB
  5. Internet

An attacker uses a compromised employee account to invoke a finance AI agent. The agent runs as its own service identity — and that identity is what reaches the data.

  • Correlate. The service identity can reach a production database with an exploitable critical CVE, customer PII and an outbound path.
  • Decide. Every condition is backed by source evidence, so policy marks the path material. If any were unknown, the path would not be treated as confirmed.
  • Approve. Suspend the service identity’s access and revoke its active token.
  • Verify. Re-read identity and cloud state and recompute the path; close it only when the path no longer resolves.

Illustrative scenario. People and systems are fictional.

Our lab proof of concept

Cross-control attack path disruption

No single control sees the whole path. Circuit builds it across endpoint, identity, cloud, network and data, enumerates the control points that can break it, has your executor act through the one you approve, and verifies that the path is gone.

Control points

  • Isolate the endpoint
  • Block the host at the subnet
  • Deny role sessions
  • Revoke sign-in sessions

You choose the cut

Circuit ranks every control point that can break the path and recommends one. Before anyone approves, any other can be used instead, or added alongside — and whoever asks for a cut never approves it.

What makes it different

Observed activity is kept separate from theoretical reachability, and every candidate action is tied to a control that can actually execute and verify it.

With Circuit AI

When an alert names a device someone is signed in on, the investigation agent asks your EDR for the device’s own connections and your cloud for the owner of the address. In our lab — where Circuit AI is scripted from the facts and calls no model — that confirmed the path 41 s after the alert, 4 min 24 s before the flow log corroborated the same connection. How the agents work

Example scenario
  1. Laptop
  2. Engineer
  3. Prod role
  4. App server
  5. Customer DB

An engineer’s laptop is compromised, and their identity can assume a production role. No single tool sees the rest of the story.

  • Correlate. Network telemetry shows the laptop talking to a production server with an exploitable critical CVE — and that server can reach the customer database.
  • Recommend. Isolate the laptop through your EDR: immediate and reversible. Blocking the host at the subnet or denying the role’s sessions can be used instead; revoking sign-in sessions can be added alongside.
  • Approve. Your approver signs the isolation with a security key.
  • Verify. Re-read the EDR and fresh network evidence and recompute the path. Remaining vulnerability and privilege risk stays open, with an owner — not silently closed.

Illustrative scenario. People and systems are fictional.

Next

AI agent runtime risk containment

AI agents use identities, credentials, tools and data. Circuit is designed to treat each agent as an enterprise principal and to evaluate requested actions against its approved scope before they execute — blocking, restricting or requiring approval when an action creates material risk.

Designed interventions

  • Block a tool call
  • Restrict scope
  • Require approval
  • Revoke runtime credentials

Deterministic by design

Allow and deny decisions are traceable to policy and evidence. If required context is missing, the decision is marked incomplete — never silently allowed. Where an agent’s tool calls have no control point of their own, a planned Circuit tool gateway provides one.

Example scenario
  1. Finance analyst
  2. Finance agent
  3. Agent identity
  4. Tool gateway
  5. Cloud IAM

A legitimate analyst uses a finance agent. Mid-session, the agent asks to create a new long-lived access key for its own identity.

  • Decide. Creating credentials is outside the agent’s approved scope — an explicitly prohibited privileged action.
  • Contain. Circuit’s planned tool gateway, an enforcement point in the agent’s tool-call path, denies the call before it reaches the cloud. A policy-explicit deny needs no human approval.
  • Escalate if needed. If the agent’s credential itself is at risk, Circuit recommends revoking it — an action that can require approval.
  • Verify. Re-read identity and cloud state to confirm that no new key exists and the agent stayed within its boundary.

Illustrative scenario. People and systems are fictional.

Who it’s for

Built for the teams who own the outcome.

Exposure management

Turn the material paths your exposure program finds into verified interruptions, not more tickets.

Security operations

Act on material paths with a clear, approved action — not another alert queue.

Identity & access

Remove the privilege, session or credential that actually creates exposure.

Cloud security

Break cloud attack paths with bounded, reversible changes that respect production.

Security leadership

Report risk reduced with verified outcomes and signed evidence, not anecdotes.

Risk, audit & compliance

Get every action with its policy, approval, response and verification attached — and a CERT-In draft from the sealed record.

The outcome

Time to verified interruption.

TTVI measures the time from the moment Circuit’s policy confirms a path is material to the final verification that it no longer resolves. It’s the number Circuit is built to shrink — and every step of it is on the record.

TTVI, step by step
  1. Path confirmedEvery condition backed by evidence
  2. Action approvedSigned by your approver
  3. Change appliedThrough your own control
  4. Verified gonePath recomputed at the source

TTVI is measured from the evidence record — not from estimates. In our lab: 9 min 24 s from the confirmed path to the verified cut.

Complements, never replaces

Where Circuit fits in your stack.

Circuit is the layer between knowing and verified action. Your existing tools stay — as sources of context and as points of enforcement.

Category

SOAR

What it does

Runs predefined workflows after an alert or case exists.

How Circuit works with it

Circuit decides whether intervention is warranted, picks a permitted action and verifies the result. SOAR can be one of the ways Circuit acts.

Category

CTEM & attack-path management

What it does

Discovers, prioritizes and validates exposure.

How Circuit works with it

Circuit adds approval-gated, verified mobilization: it turns well-understood, material exposure into controlled intervention. Their findings are valuable context.

Category

BAS & exposure validation

What it does

Simulates or emulates attacks to test whether exposure is exploitable.

How Circuit works with it

Circuit doesn’t attack your environment. It proves a path from your own tools’ evidence, breaks it through a control you approve and verifies the break. Validation results are useful context.

Category

ITDR & NHI security

What it does

Detects identity threats and inventories non-human identities.

How Circuit works with it

Their findings are context. Circuit acts through your identity provider to break the path an identity is part of, then verifies it.

Category

EDR, IAM, CNAPP & network controls

What it does

Protect and enforce within their own domain.

How Circuit works with it

Circuit coordinates them across domains — reading their context and acting through their APIs, with your approval.

Category

SIEM & security data lakes

What it does

Collect and retain security telemetry at scale.

How Circuit works with it

Circuit is designed to ingest only the context its decisions need, to avoid SIEM-scale data costs.

Which path worries you most?

Tell us where your exposure lives today. We’ll show you how Circuit is designed to interrupt it — and prove it.