Standard diagnosis of SARS-CoV-2 by nasopharyngeal swab (NPS) and real-time reverse transcriptase-polymerase chain reaction (PCR) requires a sophisticated laboratory, skilled staff, and expensive reagents that are difficult to establish and maintain in isolated, low-resource settings. In the remote setting of tropical Sumba Island, eastern Indonesia, we evaluated alternative sampling with fresh saliva (FS) and testing with colorimetric loop-medicated isothermal amplification (LAMP). Between August 2020 and May 2021, we enrolled 159 patients with suspected SARS-CoV-2 infection, of whom 75 (47%) had a positive PCR on NPS (median cycle threshold [Ct] value: 27.6, interquartile range: 12.5-37.6). PCR on FS had a sensitivity of 72.5% (50/69, 95% confidence interval [CI]: 60.4-82.5) and a specificity of 85.7% (66/77, 95% CI: 75.9-92.6), and positive (PPV) and negative (NPV) predictive values of 82.0% (95% CI: 0.0-90.6) and 77.6% (95% CI: 67.3-86.0), respectively. LAMP on NPS had a sensitivity of 68.0% (51/75, 95% CI: 56.2-78.3) and a specificity of 70.8% (63/84, 95% CI: 58.9-81.0), with PPV 70.8% (95% CI: 58.9-81.0) and NPV 72.4% (95% CI: 61.8-81.5%). LAMP on FS had a sensitivity of 62.3% (43/69, 95% CI: 49.8-73.7%) and a specificity of 72.7% (56/77, 95% CI: 61.4-82.3%), with PPV 67.2% (95% CI: 54.3-78.4) and NPV 68.3% (95% CI: 57.1-78.1%). LAMP sensitivity was higher for NPS and FS specimens with high viral loads (87.1% and 75.0% for Ct value < 26, respectively). Dried saliva on filter paper was stable for 4 days at room temperature. LAMP on either NPS or FS could offer an accessible alternative for SARS-CoV-2 diagnosis in low-resource settings, with potential for optimizing sample collection and processing, and selection of gene targets.