VorticVortic
PlatformSolutionsPricingBlogSign inRequest access
◐ Platform

Nine specialist agents. One surface. Every claim cited.

Vortic is the system of action for underwriting. The platform is built around three primitives: a Command Center that surfaces decisions, a Run Pipeline that fires nine specialists in parallel, and a credits ledger that meters every action. Below, what each piece does.

Command Center — daily view

What needs your attention today. SLA breaches at the top. Pending decisions, yesterday's closures, concentration alerts — pre-surfaced. A chat that already knows your day. Click through to bind, decline, refer, or delegate.

  • ·Contextual chip suggestions seeded from your data
  • ·Three-envelope chat (answer / navigate / task)
  • ·Per-agent credits ledger visible inline
  • ·First thing you see when you sign in

Workbench — detailed view

Case-level depth. The submission you're working on right now: full file, agent traces, decision history, learning loops, action focus. Where you go to actually understand and decide one risk.

  • ·Case file, geo context, portfolio impact, similar bound risks
  • ·Agent run history with sources cited
  • ·Decision learning events captured for the curator
  • ·Action focus bar — next step always visible

Run Pipeline — playground

Paste a submission, watch nine agents fan out in real time. SSE streaming, every chunk live. Switch between Auto, Copilot, and Full Review modes. The fastest way to test a risk shape or a model swap.

  • ·SSE streaming, sources cited inline
  • ·HITL gates after parse + analysis (review mode)
  • ·Per-phase credit pre-flight, per-agent refunds on failure
  • ·Median run on demo submissions: ~30 seconds

Nine specialist agents

Parse, Risk, Flood, Pricing, Compliance, Treaty, Portfolio, Memo, plus Triage and Orchestrator. Each agent has its own system prompt, model assignment, and trace history. Configurable per agent.

  • ·Risk · construction, occupancy, loss history
  • ·Flood · EA Flood (UK) + FEMA NFIP (US) integrations
  • ·Pricing · rate adequacy vs market benchmarks
  • ·Treaty · postcode (UK) + ZIP (US) aggregation
  • ·Memo · synthesises everything with sources cited

Agent Studio

Edit any agent's system prompt and model assignment. Test prompts in the simulator before persisting. Version history with rollback. All changes audited.

  • ·Curated free OpenRouter model registry (Qwen · DeepSeek · Gemini · OSS)
  • ·Test in simulator before persisting
  • ·Version history with rollback
  • ·Injection-pattern sanitization on every save

Data Fabric

Public flood APIs (EA Flood UK, FEMA NFIP US) wired in production today. Verisk ISO, D&B Firmographics, Lloyd's Market Data, OS Postcodes available on Enterprise tier with credentials. Inbound + outbound webhooks for broker placement integrations.

  • ·EA Flood + FEMA NFIP — live, cached for 30 days
  • ·Inbound webhook for broker submissions (Enterprise)
  • ·Outbound webhook for bind / decline events (Enterprise)
  • ·BYO data warehouse on Enterprise (Snowflake, Postgres)

Audit trail + compliance

Every LLM call logged in agent_traces with model, tokens, duration, and source citations. Every decision generates a structured memo with cited sources. Every prompt change recorded with old/new diff. Hand to compliance and they sign off.

  • ·Per-call trace rows with PII redaction at insert
  • ·Structured decision memos at /submission/[id]/history
  • ·Prompt change history with diff in Agent Studio
  • ·GDPR-compliant data handling · SOC 2 Type II prep in flight

Credits ledger

Per-agent metering — see exactly what you spend per run. Atomic spend RPC enforces balance before fan-out, refunds on individual agent failure. Pay $0.10 per credit, $20 minimum top-up via Stripe Checkout.

  • ·Pre-flight charge prevents mid-pipeline depletion
  • ·Failed agents auto-refunded
  • ·Append-only transaction ledger
  • ·Idempotent Stripe webhook handling

Bound book + renewals

Live portfolio view with concentration analysis. 60-day renewal cron auto-creates renewal submissions and emails the underwriter. CSV bulk import for new partners onboarding their existing book.

  • ·Concentration alerts by postcode + occupancy
  • ·60-day automatic renewal pipeline
  • ·CSV bulk import with dry-run preview
  • ·Per-treaty cap configuration

Orchestrator + AI Workspace

Universal prompt entry. The orchestrator routes natural-language questions to the right agents in parallel; synthesis merges results. Useful for "what's our exposure to LS9 if we bind this?" type questions.

  • ·Natural-language intent classification
  • ·Parallel multi-agent execution
  • ·Synthesis layer merges + cites
  • ·Session history persisted
◑ Stack

What it’s built on.

FrontendNext.js 16 · React Server Components · Tailwind v4
DatabaseSupabase Postgres · Row-level security
LLM gatewayOpenRouter · Llama 3.3 / DeepSeek / Qwen / Gemini · paid fallback
StreamingServer-Sent Events on every agent call
EmailResend · React Email templates
BillingStripe Checkout · idempotent webhooks · per-agent credits ledger
Rate limitsUpstash Redis · per-IP + per-user buckets · per-route concurrency caps
MonitoringSentry (errors, traces) · PostHog (events) · structured logs
CaptchaCloudflare Turnstile on access requests
DeployVercel · daily cron · region-aware

See how it fits your book.

Vortic is invite-only during early access. Tell us about your team and we’ll tune the platform to your lines.

Request access See solutions →