Tuberculous meningitis (TBM) is a devastating infection of the central nervous system lacking an adequate point-of-care diagnostic test. We conducted a prospective cohort study of 550 Zambian adults with suspected TBM to determine the diagnostic accuracy of cerebrospinal fluid (CSF) Xpert MTB/RIF, CSF lipoarabinomannan (LAM), urine LAM, CSF total protein, and CSF glucose compared with the gold standard of CSF culture. We categorized patients with a positive CSF tuberculosis (TB) culture as definite TBM. We also assessed inpatient and 1-year mortality on definite TBM patients when CSF Xpert MTB/RIF results were available in real time to treating physicians relative to a historical comparison cohort in whom Xpert results were not available in real time. Of the 550 patients, 474 (86.2%) were HIV-infected and 105/550 (19.1%) had definite TBM based on a positive CSF culture. The sensitivity/specificity of the diagnostic tests were CSF Xpert MTB/RIF, 52.9%/94.2%; CSF LAM, 21.9%/94.2%; urine LAM, 24.1%/76.1%; and CSF glucose <40 mg/dl, and total protein, >100 mg/dl, 66.3%/90%. A model including CSF Xpert MTB/RIF, CSF LAM, CSF glucose, and CSF total protein demonstrated an area under the receiver operating curve of 0.90. The inpatient and 1-year mortality for definite TBM was 43% and 57%, respectively. There was low sensitivity for the diagnosis of TBM across all diagnostics tests. CSF Xpert MTB/RIF and CSF LAM are highly specific for the diagnosis of TBM. Despite the use of Xpert MTB/RIF for diagnostic purpose in real time, TBM was still associated with a high mortality in Zambian patients.