Product case study
Live productTerrabit
Agentic infrastructure for parametric insurance.
AI agents design, price, validate, and monitor parametric insurance indices on public data. Every price, trigger, and payout comes from a deterministic engine and reproduces from a frozen record.
Hurricane Beryl, 121.5 kt at 72 km. Trigger 64 kt.
USD 39.38mPAYOUT · 78.8% OF LIMITReproduced: exact matchThe challenge
Parametric cover is only as good as its index.
Parametric insurance pays on a measured index instead of an assessed loss, so it can settle quickly, but only if the index tracks the loss it is meant to cover. Getting there takes decades of weather, water, and hazard data, backtests against real events, an honest view of basis risk, and proof years later that a trigger was computed correctly. Few insurers or brokers have that scientific infrastructure in house.
Our response
Agents that research, engines that decide.
Terrabit splits the work in two. AI agents turn a plain-language brief into candidate indices, run the reviews, watch live seasons, and explain what they find. Every contractual number comes from a deterministic, versioned engine, and every settlement reproduces from a fingerprinted record, so no agent ever has discretion over a payout. It runs as a standalone platform or embeds into existing insurance systems through a REST API, a Python SDK, and MCP.
What we built
A working product, not a concept deck.
- An index design agent that turns a one-sentence brief into priced candidate indices, scored against loss history
- 30 deterministic engines across weather, water, cyclone, earthquake, space-weather, cyber, and digital-asset perils
- Pricing with burn analysis, fitted extreme-value tails, and bootstrap intervals on every number
- A monitoring agent that re-checks every saved contract on fresh data and alerts on trigger approach, delayed feeds, and revised sources
- Trigger verification on frozen, fingerprinted inputs, an independent station cross-check for weather triggers, and a settlement report with recorded human sign-off
- Portfolio accumulation and reinsurance layer analysis, plus weekly assurance that proposes renewal terms without touching existing ones
Capabilities demonstrated
What this proves we can bring to client work.
Agentic index design
Catastrophe and climate risk pricing
Backtesting and basis-risk validation
Autonomous season monitoring
Reproducible trigger verification
API, SDK, and MCP integration
Where this applies
Have a related problem?
We can help evaluate the opportunity, design the system, and move it into production.
Book a technical fit call