Recent large-scale palaeoclimate reconstructions of past temperature have been essentially biased to the extratropics owing to a paucity of proxy data in tropical regions. Herein we describe the first coral-based reconstruction of sea surface temperatures (SSTs) for the whole of the tropics (30°N–30°S). It was developed from 14 disparate coral records located in the Indian and Pacific oceans. Over the most replicated period, the reconstruction explains 57% of the tropical SST variance. However, the strength of this signal weakens markedly as the number of coral records decreases. The reconstruction is robust between 1850 and 1993, but some fidelity is indicated back as far as the mid 18th century. These results suggest that ambiguities in the low frequency domain of δ18O measurements can be partially overcome by pooling together multiple time series from different locations around the tropics. Agreement with simulations from two general circulation models indicates that the late 20th century is likely the warmest period in the tropics for the last 250 years, and that this recent warming can only be explained by anthropogenic forcing. The high frequency variability is dominated by the El Niño–Southern Oscillation. The reconstruction, owing to the small number of coral records, is unfortunately restricted both in time and space. Therefore we hope that this study will spur the palaeoclimate community to develop new and longer proxy series to improve the current meager data-base of temperature sensitive series in the tropics.