Financial models as code. Built by your AI agent.
Describe your business to your AI agent. Flatland compiles it into a typed financial model where every assumption is named, every dependency is traced, and changing one thing doesn’t silently break another. Under 5 minutes.
Works with Claude Code · any MCP client
Models stored locally · Your data never leaves your machine
Under 5 minutes, start to finish.
You live in Flatland.
Your spreadsheet cell doesn’t know what it is.
Your versions aren’t controlled.
Your formulas break silently.
“I looked, and, behold, a new world!”
Flatland: A Romance of Many Dimensions — Edwin Abbott Abbott, 1884
Your business doesn’t live in two dimensions.
Your model doesn’t have to.
A cell that knows what it means.
A compiled model has a shape. Every assumption knows its place. Every formula traces back to typed inputs. Change one thing, and every dependency updates with it — no silent drift, no guesswork.
A scenario is an overlay, not a copy — just the drivers that change. Compile a new future instantly, without touching the base. Every assertion runs on every scenario.
Y3 EBITDA · flatland_scenario(base, upside, downside) ✓
Flatland ranks every driver by impact on any output — EBITDA, runway, LTV. One driver usually matters 10x more than the rest. Know which before you waste a week tuning something that doesn’t move.
flatland_sensitivity(target="ebitda_y3") ✓
Every assumption typed. Every dependency traced. Every change propagated instantly.
Flatland is a financial modeling engine for AI agents. Your agent connects via MCP and builds structured, typed models — not markdown tables or prose estimates, but real computation graphs with named assumptions, scenario analysis, sensitivity ranking, and named assertions that guard against invalid assumptions. Ask your agent to model SaaS unit economics, stress-test fundraising assumptions, or run a three-statement operating model. Flatland compiles it, your agent explains it, and you export a live Excel file ready for investors. Works with Claude Code, Windsurf, and any MCP-compatible tool.
I’ll build this as an integrated 3-statement model. Adding typed drivers for each statement plus the tie-out assertions.
Your 3-statement model is ready. Let me show you the full compiled output, then we can stress-test the assumptions.
I’ll create a downside scenario with those overrides.
At these assumptions, the model insolvents at month 22. Your strongest defensive lever is churn — a 1% improvement reclaims $42K of year-3 EBITDA. Margin is second. Delaying the raise is unsurvivable without churn improvement.
Want to run sensitivity on the 12 drivers most tied to cash runway?
Every cell references another. Nothing hardcoded.
Your investor can audit every number.