This paper presents an analytical system for analysis of all single substituted isotopologues (<sup>12</sup>C<sup>16</sup>O<sup>17</sup>O, <sup>12</sup>C<sup>16</sup>O<sup>18</sup>O, <sup>13</sup>C<sup>16</sup>O<sup>16</sup>O) in nanomolar quantities of CO<sub>2</sub> extracted from stratospheric air samples. CO<sub>2</sub> is separated from bulk air by gas chromatography and CO<sub>2</sub> isotope ratio measurements (ion masses 45 / 44 and 46 / 44) are performed using isotope ratio mass spectrometry (IRMS). The <sup>17</sup>O excess (Δ<sup>17</sup>O) is derived from isotope measurements on two different CO<sub>2</sub> aliquots: unmodified CO<sub>2</sub> and CO<sub>2</sub> after complete oxygen isotope exchange with cerium oxide (CeO<sub>2</sub>) at 700 °C. Thus, a single measurement of Δ<sup>17</sup>O requires two injections of 1 mL of air with a CO<sub>2</sub> mole fraction of 390 μmol mol<sup>−1</sup> at 293 K and 1 bar pressure (corresponding to 16 nmol CO<sub>2</sub> each). The required sample size (including flushing) is 2.7 mL of air. A single analysis (one pair of injections) takes 15 minutes. The analytical system is fully automated for unattended measurements over several days. The standard deviation of the <sup>17</sup>O excess analysis is 1.7‰. Multiple measurements on an air sample reduce the measurement uncertainty, as expected for the statistical standard error. Thus, the uncertainty for a group of 10 measurements is 0.58‰ for Δ <sup>17</sup>O in 2.5 h of analysis. 100 repeat analyses of one air sample decrease the standard error to 0.20‰. The instrument performance was demonstrated by measuring CO<sub>2</sub> on stratospheric air samples obtained during the EU project RECONCILE with the high-altitude aircraft Geophysica. The precision for RECONCILE data is 0.03‰ (1σ) for δ<sup>13</sup>C, 0.07‰ (1σ) for δ<sup>18</sup>O and 0.55‰ (1σ) for δ<sup>17</sup>O for a sample of 10 measurements. This is sufficient to examine stratospheric enrichments, which at altitude 33 km go up to 12‰ for δ<sup>17</sup>O and up to 8‰ for δ<sup>18</sup>O with respect to tropospheric CO<sub>2</sub> : δ<sup>17</sup>O ~ 21‰ Vienna Standard Mean Ocean Water (VSMOW), δ<sup>18</sup>O ~ 41‰ VSMOW (Lämmerzahl et al., 2002). The samples measured with our analytical technique agree with available data for stratospheric CO<sub>2</sub>.