Reading Sponda from a program
Sponda holds inflation-adjusted valuation and quality indicators for listed companies in the U.S. and Brazil. Every figure on the site is available to software four ways, listed here cheapest first.
MCP server
The best option if you are building on an assistant. Sponda is an MCP server, so the model can screen and look up companies as tool calls rather than by fetching and parsing pages.
https://sponda.capital/api/mcpStreamable HTTP, stateless, no authentication. In Claude Code:
claude mcp add --transport http sponda https://sponda.capital/api/mcp/Tools: `list_available_indicators`, `screen_companies`, `get_company`, `get_fundamentals`. Call `list_available_indicators` first; it returns the exact indicator keys, the countries and sectors present in the data, and an explicit list of metrics Sponda does not track.
Markdown pages
Every public page is also served as plain markdown at the same URL with `.md` appended. The HTML renders its numbers client-side and carries none of them; the markdown carries the table.
https://sponda.capital/en/PETR4.md
https://sponda.capital/en/PETR4/charts.md
https://sponda.capital/en/PETR4/fundamentals.md
https://sponda.capital/en/PETR4/compare.md
https://sponda.capital/en/screener.mdThe tab slug is localized: `charts` is `graficos` in Portuguese and Spanish, `graphiques` in French, `diagramme` in German, `grafici` in Italian. A company page with no locale prefix, such as `/PETR4.md`, is served in English rather than redirected, so a guessed URL still works.
Every HTML page advertises its twin twice: a `<link rel="alternate" type="text/markdown">` tag in the document, and a `Link:` response header carrying the same URL. A `HEAD` request is enough to discover it.
JSON endpoints
Public, uncapped, and the same data the markdown pages are rendered from.
GET https://sponda.capital/api/screener/?pe10_max=10&debt_to_equity_max=1&sort=-market_cap&limit=50
GET https://sponda.capital/api/tickers/{SYMBOL}/indicators/
GET https://sponda.capital/api/tickers/{SYMBOL}/indicators/?symbols=A,B,C
GET https://sponda.capital/api/tickers/symbols/
GET https://sponda.capital/api/assistant/indicators/`/api/tickers/symbols/` returns every listed company symbol and nothing else, about 150KB, which is the cheapest way to learn the universe. `/api/assistant/indicators/` returns the indicator glossary with definitions and which direction is better.
One endpoint is capped. `/api/quote/{TICKER}/` is richer than `/indicators/` but fetches from upstream data providers, so it is limited to 20 distinct companies per day per IP and answers `429` past that. Use the markdown pages or the endpoints above to read at volume.
Finding everything
https://sponda.capital/llms.txt
https://sponda.capital/sitemap.xml`llms.txt` describes the conventions on this page in a form meant to be read rather than rendered. `sitemap.xml` is an index; its children under `/sitemaps/` enumerate every company page, so nothing has to be guessed.
Blog posts are markdown too, at their permalink plus `index.md`.
What Sponda does not have
Treat any figure for these as absent rather than zero: return on equity, dividend yield, revenue growth, analyst price targets, news, and recent events. `list_available_indicators` returns the authoritative list.
One convention is load-bearing. A P/E window is empty unless the company has the full history for it, so a P/E15 is never quietly a P/E8. `pe_years_available` tells you the widest window a company can fill. An absent indicator means absent data, never zero.