Agent ready · Idle

The agent is standing by.

Pick a detected control gap below and watch the five-stage run execute in real time — policy retrieval, evidence inspection, bounded plan, human approval, validation, and filed evidence artefact.

Synthetic gaps
7

Across 5 regulated control domains

Grounded tools
5

Policy, evidence, owners, memory, blast radius

Typical run
~45s

End-to-end: diagnosis to filed evidence

Scroll for capabilities

Why this agent

Not a chatbot — a bounded remediation worker.

Grounded by policy

Every claim is cited from the internal policy library — no hallucinated regulations.

Tool-augmented

Calls retrieval tools live during the run. You see every lookup in the trace.

Memory-aware

Matches the gap against past remediations and reuses proven patterns.

Blast-radius first

Every plan lists what it touches, what could go wrong, and how to reverse it.

Detected gap queue

One click to start a run.

These are the synthetic gaps a detection platform like Brontë would surface. Select one to load the workspace with the full control context.

Full Gap Library

Identity & Access

Critical

MFA enforcement gap on privileged admin accounts

CTRL-IAM-047

Three production administrator accounts in the central identity provider remain exempt from mandatory MFA policy despite having standing access to customer, treasury, and identity administration consoles.

3 privileged administrator accountsRun agent

Data Protection

High

Encryption-at-rest gap on general ledger replica

CTRL-DP-112

A non-production replica of the general ledger PostgreSQL cluster was restored into the analytics VPC without storage-level encryption, leaving financial records exposed if the underlying volume is accessed outside the application layer.

1 database replica and 4 attached storage volumesRun agent

Change Management

High

Change approval bypass in deployment pipeline

CTRL-CHG-031

Seven production deployments in the last quarter were merged and released through the automated pipeline without the required second-engineer approval because the branch protection rule was disabled for the release service account.

7 production change recordsRun agent

Transaction Monitoring

Critical

Sanctions screening lag on cross-border payments

CTRL-TM-204

Sanctions-list updates are taking up to 36 hours to propagate from the compliance feed handler into the cross-border payments screening engine, materially exceeding the required four-hour maximum and increasing exposure to prohibited counterparties.

2 sanctions feed pipelines and 1 screening engine rule setRun agent

Vendor Risk

Medium

Vendor access review overdue for third-party support vendors

CTRL-VRM-019

Quarterly access recertification for four third-party support vendors has not been completed for two consecutive review cycles, leaving external production access active without current business-owner attestation.

4 third-party vendor profiles and 19 active entitlementsRun agent

Change Management

Critical

Unreviewed LLM used in customer complaints triage

CTRL-MDL-014

The customer complaints triage service has been promoted to production using a third-party LLM endpoint with no model-risk sign-off, no human-review gate on outbound responses, and no evidence captured for regulator-facing decisions about hardship, fees, or dispute resolution.

1 customer complaints triage microservice, 2 LLM vendor endpoints, ~4,200 monthly customer-facing responsesRun agent

Data Protection

High

Audit log retention gap on payments service

CTRL-DP-188

Audit logs for the payments microservice are being purged after 30 days in the observability pipeline instead of the required 12 months, reducing the ability to investigate fraud events and meet card scheme evidence obligations.

1 payments microservice log stream and 12 index partitionsRun agent