Financial models, built by your AI agent.

One prompt in Claude or ChatGPT. A live model your team can steer.

Use Finicast (https://finicast.com/for-agents) to build a five-year forecast for CoreWeave

Works in Claude, ChatGPT, Claude Code, Codex and Cursor. No account, no connector.

The dashboard the prompt produces: an editable drivers table, KPI tiles and charts for a five-year forecast

Here is your model in Finicast:CoreWeave five-year forecastdashboard · editable assumptions · 7 cardsOpen →

Built by Claude from that prompt. Change a driver and every statement follows. Open the live model

< 1 msto recalculate a 100,000-row ledger

12 ruleswhere a spreadsheet needs 480 formulas

1 requestfrom document to statements and a link

0 accountsto start: npm install -g finidb

From prompt to model in three steps.

The agent writes rules, not cells.

revenue[frame=fcst] = PREV(revenue) * (1 + growth)
gross_profit        = revenue - cogs
cash[frame=fcst]    = PREV(cash) + cash_flow.free_cash_flow

2
Finicast calculates every period.
Income statement, cash flow, balance sheet. One document in, statements and a link out.
3
You steer.
Change a driver. Every dependent cell updates in milliseconds, for everyone with the link.

The Finicast workspace: line items on rows, years on columns, the rule bar below

Built for agents. Made for people.

Rules, not cells

A rule names a line item and the region it covers. One rule, every forecast period.

A real dependency graph

An edit recomputes only what depends on it. Verified against a reference evaluator.

Dashboards people can edit

Drivers in a table, KPIs and charts that follow. The agent declares them in the same document.

Every interface an agent needs

MCP, REST, a JSON model document, an npm CLI, a skill file. Every error carries a fix.

Why not a spreadsheet, or a script.

	Spreadsheet	Script	Finicast
A 40-line, 12-period statement	480 cell formulas	A few hundred lines	≈ 12 rules
Forecast vs. actual periods	Hand-picked ranges	Index arithmetic	frame = fcst
One input changes	Whole workbook	Re-run the script	Only what depends on it
What the person gets	A file	A number	A live model to steer

One document in. Statements and a link out.

{
  "periods":  { "start": "2022-01", "count": 9, "grain": "year", "histUntil": "2025-12-31" },
  "pivots": {
    "assumptions":      { "lines": ["revenue_growth", "gross_margin"], "inputs": { "revenue_growth": { "fy2026": 0.6 } } },
    "income_statement": { "lines": ["revenue", "cogs", "gross_profit"],
                          "inputs": { "revenue": { "fy2025": 5131 } },
                          "rules":  "revenue[frame=fcst] = PREV(revenue) * (1 + assumptions.revenue_growth)" }
  },
  "outputs": "all",
  "dashboards": "auto"
}

curl -X POST https://finicast.com/api/build --data @model.json
# → statements as markdown + a link to the live workspace

Hosted MCP finicast.com/api/mcp
Local npx -y finidb mcp · build offline with finidb build
Source-available under PolyForm Noncommercial

Docs for agents → llms.txt

Free to learn on. Licensed for business.

Community

Free

Self-hosted from npm, for noncommercial use.

npm install -g finidb

Cloud

Free during preview

Hosted workspaces, dashboards and MCP at finicast.com.

Start with the prompt

Commercial

Per organization

A license for the engine in a business or a product.

Talk to us

Ask your assistant for a model.

Copy the prompt Open the live model