R packages

I write packages that hand back Brazilian public data tidy, straight from R

Brazilian public data rarely arrives ready to use. It comes as a fixed-width file, a spreadsheet with a three-row header, something behind a form, or an API returning nested JSON with uppercase field names. The cleanup work is always the same, and always redone by whoever arrives next.

I wrote these packages to do that work once, in R, and hand back a tidy data frame.

On CRAN

CRAN is not hosting: it is a gate with a human reviewer, checks across several platforms, and a requirement that documentation and examples actually run. I put two packages through it.

educabR

Brazilian education data from INEP — school census, IDEB, ENEM, SAEB, higher education census, ENCCEJA, IDD and ENADE. Downloads, unpacks and tidies.

install.packages("educabR")

Version 1.1.0 on CRAN since 2026-08-23 · documentation · GitHub

healthbR

Brazilian public health data from several DATASUS and Ministry of Health sources — hospital admissions, mortality, births, facilities, and surveys such as VIGITEL.

install.packages("healthbR")

Version 0.2.0 on CRAN since 2026-02-15 · documentation · GitHub

In development, on GitHub

Installable with remotes::install_github(). I have not submitted these to CRAN yet.

cvmdata

Open data from the CVM, Brazil’s securities regulator — listed companies and funds, in tidy form.

0.2.0.9000 · documentation · GitHub

welfarebR

Brazilian social assistance data — programmes, benefits and coverage.

0.1.0 · GitHub

senado

The Brazilian Federal Senate open data API, with a tidy interface. The R sibling of the Senate MCP server.

0.0.0.9000 · GitHub

Why R, and why MCP as well

The MCP servers and these packages are how I attack the same problem from opposite sides. A package serves someone already inside an analysis, writing code, who wants the data in the environment where the modelling will happen. An MCP server serves someone talking to an assistant who wants the answer with its source, without writing any code.

Neither replaces the other, and both depend on the same thing: somebody having read the source’s documentation, found where it lies, and written that down in code someone else can reuse. That is the work I do.