Agent risk
you can underwrite.
Minogan watches every AI agent your insureds run — around the clock, per company — and turns behavior, failures and drift into a living risk score your underwriters can actually price.
Prefer to talk? (305) 720-0839
Q4 renewal book
38 insureds · refreshed 4s ago
Watched agents
1,284
- CI18
Claims Intake
Harborline Mutual
- QA42
Quote Assistant
Northgate Specialty
- PS27
Policy Servicing
Meridian Casualty
- FV71
FNOL Voice Agent
Atlas Ridge P&C
Portfolio risk index
Behavior
22
Failures
41
Exposure
38
Signal feed
UTC 03:14Tool-call loop detected
nowFNOL Voice Agent · 14× retries
Hallucinated coverage term
2mQuote Assistant · flagged by policy guard
Model version changed
11mClaims Intake · model v2.2 → v2.3
Drift resolved
26mPolicy Servicing · back to baseline
Illustrative names from the specialty market we serve — carriers, MGAs and syndicates
Everything you need to insure autonomous work
One platform from first telemetry to bound policy. Built with actuaries, claims leads and AI engineers in the same room.
24/7 agent monitoring
Every prompt, tool call and hand-off streamed and evaluated in real time — nights, weekends and holiday peaks included.
Explainable risk scoring
38 weighted signals roll into one 0–100 score — with every point traceable to evidence.
Per-company tenancy
Each insured gets a sealed workspace. You get the portfolio view across all of them.
Behavior & failure tracking
Loops, refusals, hallucinated terms, silent tool errors — clustered into patterns, not noise.
Underwriting & business plans
Auto-drafted risk-improvement plans with conditions, tiers and pricing guidance your team can edit and bind.
- Human-in-loop on claims > $25k
- Retry cap ≤ 3 per tool call
- Weekly red-team on voice agent
Alerts that route themselves
Severity-aware routing to underwriting, claims or the insured's own engineers — wherever they already work.
Built for
Sub-minute
detection
From first anomalous tool call to a routed, evidence-backed alert — before the next shift starts.
Signal families
4
Score range
0–100
From telemetry to terms
in three quiet steps
Live in an afternoon for a pilot insured. Portfolio-wide in weeks, not quarters.
Connect the agents
Insureds drop in our SDK or point their OpenTelemetry exporter at Minogan. No prompts leave their cloud unless they choose.
import { watch } from "@minogan/sdk"
watch(agent, {
tenant: "harborline",
policy: "HM-2026-0412",
})
Baseline & score, 24/7
We learn each agent's normal, then score deviations, failures and exposure continuously — with evidence attached to every point.
Underwrite with confidence
Turn live scores into terms, conditions and business plans. Bind, endorse or refer — and keep watching after the policy is live.
Suggested: +6% rate, retry-cap endorsement, quarterly review.
Your whole book of AI risk, on one calm screen
Portfolio heat, agent forensics and plan drafting — the same view for underwriters, risk engineers and claims.
Portfolio / Atlas Ridge P&C/ Portfolio
Insured companies
38
+3 this month
Agents monitored
1,284
+112
Open incidents
7
2 critical
Premium at risk
$4.2M
↓ 11% QoQ
Risk heat · last 24h
Needs attention
- 71
FNOL Voice Agent
Atlas Ridge · Tool loop
- 42
Quote Assistant
Northgate · Hallucinated term
- 39
Broker Copilot
Bluewater · PII in output
- 31
Subrogation Bot
Meridian · Latency SLA
Evidence, not vibes
Every score links to the exact traces, tool calls and transcripts behind it.
Built for renewals
Compare an insured's agents period-over-period, right inside the renewal workflow.
Shareable with insureds
Give each company a scoped view so their engineers can fix what you flag.
“We stopped guessing what our insureds’ agents were doing at 3 a.m. A living risk score our underwriters can defend in committee — and a plan the insured can actually follow — is the difference between pricing AI and avoiding it.”
Design partner
Specialty carrier · cyber & tech underwriting
Score
0–100 living risk index
Tenants
Isolated per insured
Plans
Auto-drafted, editable
Coverage
24/7 signal monitoring
Illustrative quote from early design conversations — not a published case study.
Carrier-grade by design
Built for insurers who handle sensitive insured operational data — with isolation, redaction and auditability as defaults.
Hard tenant isolation
Per-company encryption keys and logically separated data planes. No cross-tenant queries, ever.
Your data stays home
Run the collector in your insured's VPC. Redact prompts at the edge; send only signals.
SSO, SCIM & roles
Okta, Entra ID and Google. Underwriter, risk engineer, insured and auditor roles out of the box.
Immutable audit trail
Every score change, plan edit and export is signed and retained for regulatory review.
Priced like a tool, valued like a loss ratio
Start small with one program. Scale as your AI book grows.
Starter
For MGAs piloting AI-agent cover with a handful of insureds.
billed annually
Start a pilot- Up to 5 insured companies
- 100 monitored agents
- Daily risk scores
- Email & Slack alerts
- 30-day evidence retention
Growth
For carriers writing AI exposure across a growing book.
billed annually
Book a demo- Up to 50 insured companies
- 1,500 monitored agents
- Real-time 24/7 scoring
- Auto-drafted underwriting plans
- Insured-facing portals
- SSO & 1-year retention
Enterprise
For global carriers, reinsurers and Lloyd's syndicates.
volume & program-based
Talk to sales- Unlimited tenants & agents
- Private cloud or on-prem collector
- Custom scoring models
- Portfolio & accumulation analytics
- Dedicated risk engineer
- Custom SLA & support
Questions underwriters ask us
Can't find what you need? Call us at (305) 720-0839 or book a walkthrough.
Commercial insurers, MGAs, reinsurers and Lloyd's syndicates that underwrite companies running AI agents in production — customer service, claims, sales, operations and internal copilots.
24/7 clarity on every AI agent you insure.
See your portfolio scored live in a 20-minute walkthrough — using a sample book that looks like yours.