Murata makes one I use with good results. And it has an output proportional to the angular velocity (or angular rate) so that if you want to measure angles you should integrate the signal.
The type depends on the maximal angular rate of the device you find from 75 to 300 (500)°/s.
Does signal integrating cause an error accumulation due to finite accuracy for initial conditions? Just curious to know. I do think of my angle measuring project based on load cell.
Of course you can if possible use load cells or acceleration sensors (sensing "g") to measure an angle with functions arcsine and arccos. But it is only possible if the axis is horizontal. With an axis not horizontal it is not possible to measure an unlimited angle without using an encoding device. The use of a gyroscope gives a lot of freedom but of course requires a signal processing.
I am interested in uncertainty less 0.5% and with simple algorithms for integration it is possible to reach it. With new devices speeds even over 60rpm can be measured without a fixed reference which is for some applications of great interest.