Making budget data easier to interrogate
Federal budget projections are spread across releases, workbooks, and program tables. Finding a number often means first finding the right file, understanding its structure, and keeping track of the release being compared.
CBO Data Explorer brings a conversational interface to that work. It connects questions about program spending, economic projections, and budget totals to structured data tools.
How it works
The project combines Python, DuckDB, and Model Context Protocol tools. A language model interprets the question and selects a data function; the function retrieves or calculates the result. The underlying data includes program-level baseline workbooks and the official US-CBO data repository.
This separation makes the calculations inspectable. It also supports both a browser interface and a command-line workflow.
Explore the project
Try a question about projected spending for a program, or compare values across years. The repository documents the data preparation pipeline, available tools, example questions, and tests.
Browse the source and documentation. For the upstream workbook-to-CSV pipeline, see Data Friendly CBO Baseline Detail.