Quickstart

Up and running
in 5 minutes.

Your AI agent builds the model. Flatland compiles it.

01

Prerequisites

You need an AI coding agent installed on your machine. Flatland works with:

Claude Code Cursor Windsurf

You also need Node.js installed (any recent version). That's it.

02

Subscribe and get your API key

Subscribe at flatlandfi.com. Your API key arrives by email within seconds. Keep it — you'll need it in the next step.

The key looks like: fl_live_xxxxxxxxxxxxxxxxxxxx

03

Connect your agent

Run this command in your terminal, replacing <your-key> with the key from your email:

$ npx flatland-setup <your-key>
Configuring Claude Code...
✓ Flatland MCP server registered
✓ API key saved
✓ Ready. Open your agent and start modeling.

Configures Claude Code, Cursor, or Windsurf automatically based on what's installed.

04

Describe your business

Open your agent. Use this prompt to build your first model:

Starter prompt
Build me a 3-year P&L for a SaaS doing $12K MRR with 8% monthly churn and $800 CAC. Include gross margin at 78%. Add assertions to flag if EBITDA goes negative or churn exceeds 12%.

You can describe your own business instead — the more specific you are, the more accurate the model. Revenue, costs, growth rate, margins. Whatever you know.

05

What happens next

Your agent calls Flatland's tools to build and validate the model:

  • Initializes a new model with typed drivers
  • Resolves the dependency graph — which outputs depend on which assumptions
  • Compiles: evaluates every computed value in order
  • Runs your assertions — flags anything that violates your conditions
  • Returns the compiled results: revenue, gross profit, EBITDA across all three years
flatland_compile()
Revenue   Y1: $144,000   Y2: $264,960   Y3: $487,526
Gross Profit   Y1: $112,320   (78%)
EBITDA   Y1: $67,320   Y2: $161,669   Y3: $335,270
5 assertions passed. 0 warnings.
06

What to try next

Once your base model is compiled, ask your agent to go further:

Run scenarios — base, upside, downside Sensitivity analysis — which assumptions matter most? Add more drivers — headcount, CAC, payback period Save the model — load it again in a future session Compare scenarios — what changes between upside and downside?

Just describe what you want in plain language. Your agent handles the tool calls.