With the increased sensitivity of modern nuclear magnetic resonance (NMR) spectrometers, the minimum amount needed for chemical-shift referencing of NMR spectra has decreased to a point where a few microliters can be sufficient to observe a reference signal. The reduction in the amount of required reference material is the basis for the NMR Capillary-tube Package (CapPack) platform that utilizes capillary tubes with inner diameters smaller than 150 µm as NMR-tube inserts for external reference standards. It is shown how commercially available electrophoresis capillary tubes with outer diameters of 360 µm are filled with reference liquids or solutions and then permanently sealed by the arc discharge plasma of a commercially available fusion splicer normally employed for joining optical fibers. The permanently sealed capillaries can be used as external references for chemical-shift, signal-to-noise, resolution, and concentration calibration. Combining a number of permanently sealed capillaries to form CapPack devices leads to additional applications such as performance evaluation of NMR spectrometers and NMR pulse sequences. A 10-capillary-tube side-by-side Gradient CapPack device is used in combination with one or two constant gradients, produced by room-temperature shim coils, to monitor the excitation profiles of shaped pulses. One example illustrates the performance of hyperbolic secant (sech) pulses in the EXponentially Converging Eradication Pulse Train (EXCEPT) solvent suppression sequence. The excitation profile of the pulse sequence is obtained in a single gradient NMR experiment. A clustered T1 CapPack device is introduced consisting of a coaxial NMR-tube insert that holds seven capillary tubes filled with aqueous solutions of different concentrations of the paramagnetic relaxation agent copper(ii) sulfate (CuSO4). The different CuSO4 concentrations lead to spin-lattice relaxation times in the seven capillary tubes that cover a range which extends to more than an order of magnitude. Clustered T1 CapPack devices are best suited to quantify the effects that relaxation has on magnetizations and coherences during the execution of NMR experiments, which is demonstrated for the order-of-magnitude T1 insensitivity of signal suppression with EXCEPT.