Cornerstone — guide

Cornerstone is a reinsurance pricing demo you can open and explore. It models the work of a property reinsurance underwriter: an insured company asks for cover, each underwriting year gets a pricing, the pricing carries the schedule of insured property risks, and the reinsurance structure — layers and sections — is laid on top. Insureds and pricings are MeshWeaver nodes, so they appear in search and the chat can open them. The property risks, layers and sections are not separate nodes: they are typed data collections inside each pricing, which the pricing's views read.

It is meant for insurance and reinsurance teams evaluating MeshWeaver, and for developers who want a complete, realistic example of a domain modelled as node types with typed content, data grids, a map and in-mesh tests.

Key concepts

The space holds three node types, compiled live from their source:

• Insured (Cornerstone/Insured) — a company seeking cover: name, description, industry, location, address, website and contact email. Its page is a pricing catalog that lists the insured's pricings grouped by status, each shown as a card.
• Pricing (Cornerstone/Pricing) — one underwriting year of one insured: inception and expiration dates, underwriting year, line of business, country, currency, legal entity, broker, primary insurer and a status (a new pricing starts as Draft). Pricings live beneath their insured, one per year.
• Article (Cornerstone/Article) — a documentation article with title, authors, tags, publication date, status (draft, published, archived) and a Markdown body. The space's own documentation is written as articles.

Inside a pricing sit three data collections. They are loaded from the space's sample files when the pricing is first opened — so those files must be present — and only the fullest sample pricing has them; the other pricings open with empty collections:

• Property risks — one row per insured location, with address, city, state, county, postal code and country; the sums insured for building, contents and business interruption, each with its currency; occupancy and construction codes, build and upgrade year, number of stories, sprinklers; and the geocoded latitude and longitude. Each risk remembers the source file and row it was imported from.
• Reinsurance acceptances (layers) — cession and share, estimated premium income (EPI), rate, commission, brokerage, tax, reinstatement premium, no-claims bonus, profit commission and minimum and deposit premium.
• Reinsurance sections — per layer and line of business: attachment and limit, plus optional aggregate attachment and aggregate limit.

The pricing views

Every pricing has a toolbar with five views:

1. Overview — the pricing's header: insured, status, coverage period and the key terms.
2. Property Risks — the risk schedule as a data grid: location, address, city, state, country, the three sums insured, currency and coordinates.
3. Risk Map — the geocoded risks drawn on an interactive map, so concentrations of value are visible at a glance. The map needs the Maps package and a map renderer; a pricing with no geocoded risks says so instead of drawing an empty map.
4. Structure — the reinsurance program: the layers in one grid, their sections in another.
5. Submission — a file browser over the pricing's submission documents, such as the broker's schedule and the placement slip.

The sample's reinsurance structure comes from a placement slip written as Markdown: a parser extracts the EPI, the brokerage and the layer and section definitions when the pricing's data is first loaded. It is an initial sample-data loader, not a live link: editing the slip afterwards does not refresh a Structure view that has already loaded.

A typical walk-through

1. Open the Cornerstone space and pick an insured from the catalog.
2. Open a pricing; start on Overview to see the terms.
3. Switch to Property Risks to inspect the schedule, then to Risk Map to see where the value sits geographically.
4. Open Structure to read the layers and sections, and Submission for the documents behind them.
5. Ask the chat about the insureds and pricings — the agents read those nodes. The risk and structure rows live in the pricing's collections and are shown by its views, not as nodes the chat or search can find.

What you get

• Six sample insureds with yearly pricings; the fullest pricing loads more than fifty geocoded property risks and a complete reinsurance structure from its sample files.
• The three node types with their views, compiled live on your mesh, and in-mesh tests for each, run from each type's Tests area.
• Documentation articles on the architecture, references, AI agent integration and testing.

Cornerstone is free to subscribe to and requires the Maps package for the risk map.