Brazilian company search gets MCP

A Brazilian project called radar-cnpj-mcp lets an AI agent or command-line tool search the federal company registry by business activity and location. It uses CNAE, Brazil’s activity classification, and can find, count, compare and export business records. The project’s GitHub page shows 81 stars, a small sign of early interest in country-specific business data for AI agents. The project’s GitHub page is in Portuguese.

For an operator, the practical value is in turning a large public dataset into targeted questions: how many businesses in a given sector are in a city, or which local companies match a particular activity. The data is downloaded from the Receita Federal and kept on the user’s machine, rather than hosted by the project. There’s a trade-off: getting the data takes setup, and if an AI agent runs in the cloud, search results are sent to that agent’s provider. The project also warns that contact details for MEIs may belong to individuals, and users are responsible for complying with the LGPD.

This is the kind of thing we build into an AI command centre: a tool that gives an agent access to a useful, specific source of business data, with clear limits on where the data lives and how it’s used.