A novel concept has emerged whereby the higher-order self-assembly of proteins provides a simple and robust mechanism for signal amplification. This appears to be a universal signalling mechanism within the innate immune system, where the recognition of pathogens or danger-associated molecular patterns need to trigger a strong, binary response within cells. Previously, multiple structural studies have been limited to single domains, expressed and assembled at high protein concentrations. We therefore set out to develop new in vitro strategies to characterise the behaviour of full-length proteins at physiological levels. In this study we focus on the adaptor protein MyD88, which contains two domains with different self-assembly properties: a TIR domain that can polymerise similarly to the TIR domain of Mal, and a Death Domain that has been shown to oligomerise with helical symmetry in the Myddosome complex. To visualize the behaviour of full-length MyD88 without purification steps, we use single-molecule fluorescence coupled to eukaryotic cell-free protein expression. These experiments demonstrate that at low protein concentration, only full-length MyD88 forms prion-like polymers. We also demonstrate that the metastability of MyD88 polymerisation creates the perfect binary response required in innate signalling: the system is silenced at normal concentrations but upstream signalling creates a “seed” that triggers polymerisation and amplification of the response. These findings pushed us to re-interpret the role of polymerisation in MyD88-related diseases and we studied the impact of disease-associated point mutations L93P, R196C and L252P/L265P at the molecular level. We discovered that all mutations completely block the ability of MyD88 to polymerise. We also confirm that L252P, a gain-of-function mutation, allows the MyD88 mutant to form extremely stable oligomers, even when expressed at low nanomolar concentrations. Thus, our results are consistent with and greatly add to the findings on the Myddosomes digital ‘all-or-none’ responses and the behaviour of the oncogenic mutation of MyD88.