Goodness-of-fit testing is a classical statistical tool that has been extensively investigated in the batch setting, where the sample size is determined before data collection. However, practitioners often prefer procedures that adapt to a problem's complexity over those that fix the sample size in advance. Ideally, such procedures should (a) conclude early on easy tasks (and late on hard tasks), thereby using available data resources efficiently, and (b) continuously monitor the data and efficiently incorporate statistical evidence, while controlling the false discovery rate. Generally, classical batch tests are not tailored for streaming data, as valid inference after data peeking requires multiple testing corrections, leading to diminished statistical power. Following the principle of testing by betting, we design Sequential Kernel Goodness-of-fit Tests (SKGTs) to address these limitations. We perform experiments to demonstrate our sequential test's adaptability to a problem's unknown difficulty level while controlling type-I errors.