FinDash API and MCP server

The FinDash API

Base address https://findash.tech/api/v1. Requests and responses are JSON. No authentication: every endpoint is open, and browsers may call it from any site (CORS). The OpenAPI description lists every field.

Analyse monthly figures

Send each month's revenue, its spending budget (the most planned to be spent, not a sales target) and its costs, by line. The fourteen cost lines are costOfGoodsSold, labor, employeeMeals and payrollTax (direct costs); packaging, creditCardFees, marketing and hardware (operating costs); utilities and maintenance (variable overhead); rent, insurance, businessLicense and telephone (fixed overhead). Leave out a line with nothing in it, and put a cost with no line of its own in the closest one.

curl -sS https://findash.tech/api/v1/analyze \
  -H 'Content-Type: application/json' \
  -d '{"currency": "USD", "months": [{"month": "July", "year": 2026, "revenue": 72000, "budget": 66000,
       "costs": {"costOfGoodsSold": 26000, "labor": 12000, "rent": 6000, "utilities": 2800, "marketing": 2200}}]}'

The answer has every month with its fourteen cost lines (zero where none were sent), and a total across the months:

{
  "currency": "USD",
  "months": [{
    "label": "July 2026", "month": 7, "year": 2026,
    "revenue": 72000, "budget": 66000,
    "costs": { "costOfGoodsSold": 26000, "labor": 12000, "rent": 6000, ... },
    "costGroups": { "directCosts": 38000, "operatingCosts": 2200, "variableOverhead": 2800, "fixedOverhead": 6000 },
    "totalCosts": 49000, "grossProfit": 34000, "netProfit": 23000,
    "netMarginPercent": 31.94, "budgetUsedPercent": 74.24, "overBudgetBy": 0
  }],
  "total": { "label": "July 2026", "months": 1, "revenue": 72000, "netProfit": 23000, ... },
  "needsAttention": []
}

needsAttention lists the months that made a loss or went over budget, the two things the dashboard puts first. The arithmetic is the dashboard's own, so the figures match what an owner sees on the dashboard.

Errors

Every error is JSON, with an HTTP status to match: {"error": {"code", "message", "hint", "docs"}}, and for figures that can't be read, a problems list naming each field to fix, such as months[0].costs.wages.

Limits and fair use

• Up to 120 months per request.
• Figures are worked out and sent back, never stored or logged.
• The API doesn't read invoice photos, write advice or save figures to an account: those stay on the site, for logged-in owners. The terms apply.
• There is no OAuth: nothing in the API belongs to an account, so there is nothing to authorise.

MCP servers

Two Model Context Protocol servers, over Streamable HTTP, with no authentication:

https://findash.tech/api/mcp

The product. Tools: analyze_monthly_figures (the same job as the API), get_spreadsheet_template and explain_term.

https://findash.tech/api/mcp/docs

The documentation. Tools: list_pages, read_page and search_docs, over every page on this site.

To add the product server to Claude Code, for example:

claude mcp add --transport http findash https://findash.tech/api/mcp

Each server has an MCP Server Card at its address plus /server-card, and both are listed in the AI Catalog.

Tools on the dashboard page (WebMCP)

In a browser that supports WebMCP, the dashboard offers agents four tools of its own: load_monthly_figures shows an owner's figures on the page (kept in their browser, as a spreadsheet would be), read_dashboard reads back what the page shows, show_month picks the month and period on screen, and download_template saves the spreadsheet template.

Reading the site

• llms.txt: what FinDash is, when to use it and every page, with smaller indexes for the help, the dashboard and this page.
• Every page as Markdown: ask with Accept: text/markdown, or add .md to its address (/developers.md; the home page is /index.md). ?mode=agent on any page answers with llms.txt.
• The agent skill for working with FinDash, the API catalog (RFC 9727), the sitemap and the site's structured data as a schema feed.

Source code

FinDash's code is public on GitHub, with an AGENTS.md for coding agents and the agent skill in docs/skills/findash. Questions: support.findash@findash.tech.