Monsure is the configuration layer your core was missing. Describe a product once. Monsure ships the REST APIs, OpenAPI spec, TypeScript SDK, customer buy-flow, chat copilot, embeddable widget, and MCP server for AI agents. All from the same definition. It plugs into your existing core through six adapter shapes, so your AS/400 stays where it is and your customer master is never duplicated.
A versioned, immutable Product DSL describes everything: lifecycle, coverages, sum-insured ranges, parametric triggers, eligibility, compliance metadata. Edit it in JSON or by talking to the AI. Drafts diff against published — preview-impact replays the last 30 quotes through your changes before you ship.
Publishing a product spawns ten REST endpoints, an OpenAPI 3.1 document, a TypeScript SDK, and a polished docs bundle (README, Postman collection, request/response samples, PDF reference). Hand it to your tech team. Integration starts the same afternoon.
The same configured product drives a modern customer buy-flow, a per-tenant configurable chat copilot, a drop-in iframe widget for partners, and an MCP server that lets Claude (or any agent) quote, bind, and claim natively. Configure once, sell on every surface.
Edit the product spec — your APIs, OpenAPI, SDK, customer buy-flow, chat copilot, partner widget, and agent surface all update together. The carrier never has to remember to ship the change to four places.
AI Authoring: describe a new product in plain English — Monsure drafts the entire Product DSL, with coverages, eligibility, rate references, and compliance metadata, and hands it to you as an editable draft. Editor copilot: on any published product, propose changes by typing — "raise Fire to $200k, add a Cyber Response rider" — and see the diff before you save.


In-force policies stay on their bound version forever. Edits create new versions; old policies don't change.
Replay the last 30 priced quotes through your draft. Premium-delta histogram before you ship.
Every save runs full Zod schema + linter. Compliance fields (MAS, AI-Act risk) are first-class.
The dynamic router reads your published spec and serves quote, accept, bind, get-policy, aggregation, cancel, OpenAPI, SDK, and DSL endpoints — under a versioned namespace per product. No manual route registration. No integration team waiting to write connectors.

The buyer chooses the channel. Monsure makes sure every channel sees the same up-to-date product, with the same rating, the same UW rules, the same disclosures.
Six-step wizard with a sticky live-quote rail. Presets, sliders, conditional sub-fields. Fintech-grade UX — not the eleven-step single-column form your last carrier portal shipped.

Carrier authors a copilot in plain English. Persona, scope (which products), greeting, disclaimers — drafted by AI in under a minute, published in two clicks. Customer types instead of clicking.

Eight lines of HTML, postMessage events, theme via URL params. A bank, broker, or aggregator embeds your product in their dashboard. You stay carrier-of-record; they keep the user relationship.

Drop our MCP server into Claude Desktop's config — Claude can list products, quote, bind, claim, fire triggers, read OpenAPI. The same backend that powers your apps now speaks the agentic web.
> claude
claude › use forge mcp to quote a 7-day Tokyo trip
for a 35-year-old from Singapore, and bind
if the price is under S$30.
⏺ list_products → travel-lite, motor-weekend,
biz-shield
⏺ get_product_spec(travel-lite) → DSL …
⏺ create_quote → S$24.75 · UW: accept
⏺ accept_quote, bind_quote
⇒ Policy TR-2026-00FN1J · in force.Other middleware vendors are duct-taping ChatGPT onto policy admin systems built in 1998. Monsure was designed around AI from day one — every step from authoring to claims has an AI surface, with cost attribution, tenant-scoped guardrails, and full audit. One LLM gateway, one cost ledger, one policy.
Plain-English description → full Product DSL drafted, schema-validated, ready to refine. Surface used during product ideation.
On any draft version, type a change request. The model proposes a new spec; you see the diff and apply with one click.
Carrier-authored chat assistants per product line. Drafted from a paragraph, scoped to an allowlist, published with brand + persona.
Buyer types one sentence about their business; the form pre-fills industry, headcount, wages, premises hints, suggested coverages.
Claude Desktop and any MCP-aware client can natively quote, bind, claim, and fire triggers. The carrier's existing API now speaks the agentic web.
MAS / PDPA / AI-Act review of customer copy, with severity grading. Plugged into the same audited LLM gateway with cost attribution.
{
"mcpServers": {
"monsure": {
"command": "npx",
"args": ["-y", "@monsure/mcp"],
"env": {
"MONSURE_TENANT_SLUG": "demo-carrier-sg"
}
}
}
}Ten tools exposed: list_products, create_quote, accept_quote, bind_quote, fire_trigger, get_openapi — the entire platform, addressable by an agent.
Wire any data source — flight feeds, Visa MCC merchant data, weather, telematics — to a coverage. When the predicate fires, Monsure auto-pays the policy, books the claim, syncs the carrier core via the outbox. Customer's phone vibrates with their pay-out before they think to file a claim.
Carriers don't throw away the policy admin system that runs the business. Monsure connects through six pluggable adapter shapes — REST, SOAP, DB2/AS-400, IBM MQ, SFTP/EBCDIC, CDC — and a Mock for demo. An outbox + reconciliation pattern guarantees the core stays in sync without tight coupling.

We hold an opaque CustomerRef + categorical viewmodels (age band, residency, KYC tier). NRIC, names, addresses, claims history outside Monsure — stays in your core.
From the product manager configuring a parametric BI rider, to the engineering team consuming the auto-generated SDK, to the agent ringing the API on the customer's behalf — each persona gets a surface that matches the way they want to work.
Configure, version, edit, and ship products without engineering tickets. AI authoring, draft + diff + impact preview before publish.
Auto-generated REST + OpenAPI 3.1 + TypeScript SDK per product. Drop-in iframe widget. MCP server for AI agents.
MAS / PDPA / AI-Act-aware DSL fields. Immutable versioning. Per-tenant Postgres RLS. Full audit log of LLM calls with cost attribution.
Drop-in widget, theme via URL params, postMessage events. Stay carrier-of-record; partners keep the customer relationship.
Modern multi-step buy-flow with a sticky live-quote rail, a chat copilot if they prefer typing, or a partner widget on a site they already use.
Claude Desktop. Internal AI agents. Partner bots. The MCP server exposes the platform natively — quote, bind, claim, all addressable.
Most middleware in the market traces back to architectures locked in a decade ago — composable core systems, integration buses, point-and-click product builders. They do their jobs. We made different choices because the world that surrounds them is different now.
| Dimension | Most middleware | Monsure |
|---|---|---|
| Time-to-first-product | 12–24 months | Days |
| New product → live API | Manual integration sprint | Auto-generated on publish |
| Customer surfaces | Forms only | Form · chat · widget · agent |
| AI surfaces | Add-on, retrofitted | Six native; one gateway, one ledger |
| Edit a product | Engineering ticket | Edit JSON · or talk to it |
| Multi-tenant isolation | Schema-per-tenant | Postgres RLS, one DB, FORCE |
| Customer PII | Duplicated in middleware | Stays in your core |
| Agentic clients (Claude, etc.) | Not addressable | Native via MCP |
30 minutes, with one of your hypothetical products configured live on the call. We'll show every surface — form, chat, widget, agent, parametric — in one demo.