BACKGROUND: Retention of patients in HIV care is crucial to ensure timely treatment initiation, viral suppression, and to avert AIDS-related deaths. We did a randomised trial to determine whether a text-messaging intervention improved retention during the first year of HIV care. METHODS: This unmasked, randomised parallel-group study was done at two clinics in informal settlements in Nairobi, Kenya. Eligible participants were aged 18 years or older, HIV-positive, had their own mobile phone or access to one, and were able to use simple text messaging (or have somebody who could text message on their behalf). Participants were randomly assigned (1:1), with random block sizes of 2, 4, and 6, to the intervention or control group. Participants in the intervention group received a weekly text message from the automated WelTel service for 1 year and were asked to respond within 48 h. Participants in the control group did not receive text messages. Participants in both groups received usual care, which comprised psychosocial support and counselling; patient education; CD4 cell count; treatment; screening for tuberculosis, opportunistic infections, and sexually transmitted infections; prevention of mother-to-child transmission and family planning services; and up to two telephone calls for missed appointments. The primary outcome was retention in care at 12 months (ie, clinic attendance 10-14 months after the first visit). Participants who did not attend this 12-month appointment were traced, and we considered as retained those who were confirmed to be active in care elsewhere. The data analyst and clinic staff were masked to the group assignment, whereas participants and research nurses were not. We analysed the intention-to-treat population. This trial is registered with ClinicalTrials.gov, number NCT01630304. FINDINGS: Between April 4, 2013, and June 4, 2015, we screened 1068 individuals, of whom 700 were recruited. 349 people were allocated to the intervention group and 351 to the control group. Participants were followed up for a median of 55 weeks (IQR 51-60). At 12 months, 277 (79%) of 349 participants in the intervention group were retained, compared with 285 (81%) of 351 participants in the control group (risk ratio 0·98, 95% CI 0·91-1·05; p=0·54). There was one mild adverse event related to the intervention, a domestic dispute that occurred when a participant's partner became suspicious of the weekly messages and follow-up calls. INTERPRETATION: This weekly text-messaging service did not improve retention of people in early HIV care. The intervention might have a modest role in improving self-perceived health-related quality of life in individuals in HIV care in similar settings. FUNDING: National Institutes of Health and Canadian Institutes of Health Research Canadian HIV Trials Network.