BACKGROUND: WHO recommends community-wide, systematic tuberculosis screening in high-prevalence settings. C-reactive protein has been proposed as a tuberculosis screening tool for people living with HIV. We aimed to assess the performance of a point-of-care C-reactive protein test for tuberculosis screening in the community in two countries with a high tuberculosis burden. METHODS: We conducted a prospective, cross-sectional study in four communities in Zambia and South Africa, nested in a tuberculosis prevalence survey. We included adults (aged ≥15 years) who were sputum-eligible (tuberculosis-suggestive symptoms or computer-aided-detection score ≥40 on chest x-ray) and whose sputum was tested with Xpert Ultra and liquid culture. A 5% random sample of individuals who were non-sputum-eligible was also included. We calculated sensitivity and specificity of point-of-care C-reactive protein testing, alone and combined with symptom screening, to detect tuberculosis in participants who were sputum-eligible, compared with a microbiological reference standard (positive result in Xpert Ultra, culture, or both). FINDINGS: Between Feb 19 and Aug 11, 2019, 9588 participants were enrolled in the tuberculosis prevalence study, 1588 of whom had C-reactive protein testing and received results (875 [55·1%] were women and girls, 713 [44·9%] were men and boys, 1317 [82·9%] were sputum-eligible, and 271 [17·1%] were non-sputum-eligible). Among participants who were sputum-eligible, we identified 76 individuals with tuberculosis, of whom 25 were living with HIV. Sensitivity of point-of-care C-reactive protein testing with a cutoff point of 5 mg/L or more was 50·0% (38/76, 95% CI 38·3-61·7) and specificity was 72·3% (890/1231, 69·7-74·8). Point-of-care C-reactive protein combined in parallel with symptom screening had higher sensitivity than symptom screening alone (60·5% [46/76, 95% CI 48·6-71·6] vs 34·2% [26/76, 23·7-46·0]). Specificity of point-of-care C-reactive protein combined in parallel with symptom screening was 51·7% (636/1231, 95% CI 48·8-54·5) versus 70·5% (868/1231, 67·9-73·0) with symptom screening alone. Similarly, in people living with HIV, sensitivity of point-of-care C-reactive protein combined with symptom screening was 72·0% (18/25, 95% CI 50·6-87·9) and that of symptom screening alone was 36·0% (9/25, 18·0-57·5). Specificity of point-of-care C-reactive protein testing combined in parallel with symptom screening in people living with HIV was 47·0% (118/251, 95% CI 40·7-53·4) versus 72·1% (181/251, 66·1-77·6) with symptom screening alone. INTERPRETATION: Point-of-care C-reactive protein testing alone does not meet the 90% sensitivity stipulated by WHO's target product profile for desirable characteristics for screening tests for detecting tuberculosis. However, combined with symptom screening, it might improve identification of individuals with tuberculosis in communities with high prevalence, and might be particularly useful where other recommended tools, such as chest x-ray, might not be readily available. FUNDING: European and Developing Countries Clinical Trials Partnership.