Scholars interested in legislative studies typically use observable outcomes such as roll-call votes (RCVs), speeches, and campaign contributions to quantify latent aspects of legislative behavior, such as ideal points, party unity levels, and polarization. However, the value of this strategy depends on how these outcomes are obtained, which requires understanding how legislators' decision-making processes could impact the interpretations of these measures. This project introduces a novel legislative activity database of the US Congress, integrating votes, speeches, and contributions data with additional detailed information about legislative activity. Notably, our database includes the entire universe of legislative actions and identifies the legislators associated with these actions. Moreover, our new data serve as a building block to systematically examine, for instance, the data-generating processes of RCVs and speeches or questions involving legislative dynamics such as logrolling or the evolution of cosponsorships. We validate our data using existing sources of legislative activity. To demonstrate the database’s value, we briefly explore three aspects of legislative activity scholars cannot analyze comprehensively with existing data: the prevalence of RCVs, speech givers' characteristics, and the timeline of legislative actions. We find that voice votes and unanimous consent are the most frequently used voting procedures; the frequency and length of speeches are influenced by which party controls Congress, and the speech length increases inversely with the party distance in voting shares.