# Overview

> What a Layerz model is, who builds it and who reads it, and how the docs are organised for people and for AI agents.

Layerz is a financial modelling platform built for a split of work: an AI agent builds and maintains the model, a person reviews it. The agent works through the [MCP server](/docs/mcp/overview) or the [REST API](/docs/api/overview); the person works in the web app and in the Excel export. Both see the same model, computed by the same engine.

## A model in four layers

A spreadsheet mixes logic, inputs, layout and charts in the same cells. A Layerz model keeps them apart:

| Layer        | What it holds                                           | Where to read more                                                                                                       |
| ------------ | ------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| Logic        | Formulas, time lags, the dependency graph between lines | [Values and formulas](/docs/concepts/values-and-formulas), [Formula language](/docs/reference/formulas)                  |
| Structure    | Sections, lines and their roles, lists, timelines       | [Items and roles](/docs/concepts/items-and-roles), [Model and time](/docs/concepts/model-and-time)                       |
| Data         | Inputs, scenarios, actuals imported from sources        | [Scenarios and actuals](/docs/concepts/scenarios-and-actuals), [Sources and mapping](/docs/concepts/sources-and-mapping) |
| Presentation | Dashboards, charts, KPIs, number formats                | [Dashboards](/docs/concepts/dashboards), [Number formats](/docs/reference/number-formats)                                |

Because the layers are separate, an agent can change an assumption without touching a formula, a scenario never overwrites the plan it varies, and every number on a dashboard traces back to the line and the input that produced it.

## Principles

- **Typed lines.** Every line has a role: an input, a formula, a stock rolled forward, a total. Nobody guesses what a cell is.
- **Names, not cell addresses.** A formula reads `Revenue * Gross margin`, never `=C12*D4`.
- **Time is built in.** A line holds one value per period of its timeline; lags such as `Cash_M-1` replace hand-built offsets.
- **Every write is a version.** Each change by a person or an agent is recorded, comparable and restorable.
- **The context travels with the model.** Each model carries a [FINANCE.md](/docs/reference/finance-md) (currency, units, sign convention, glossary) and each workspace its conventions, read by every agent before it writes.
- **Excel stays the exchange format.** The export is the web view with live formulas; an existing workbook imports into a model.

## How these docs are organised

- **Get started**: connect an agent over MCP, or call the API.
- **Concepts**: the financial ideas behind the model, explained once for people and agents.
- **Reference**: the model format, the formula language, the number formats, the FINANCE.md fields.
- **MCP server**: how an agent works a model, and every tool with its parameters.
- **REST API**: authentication, limits and every endpoint.

Every page is also served as plain Markdown: add `.md` to its address (for example `/docs/concepts/stocks-and-flows.md`). The index for agents is [`/llms.txt`](/llms.txt).
