Yeah, but OP isn't doing RS-232
Then why does he ask about TX/RX level shifters?
He was talking about 3.3/5v logic shifting I think, not RS232 levels.
Exactly. I believe he's using a 3.3V MCU and hasn't realized that on a
FT232R you power VCCIO with your desired logic high level to set the logic levels of its serial side.
VCCIO must be between 1.8V and VCC and if there are multiple power domains in the system, must be derived from the same source as FT232R VCC, as it must *NEVER* be powered when VCC is not. For a 3.3V target and a USB bus powered FT232R , these constraints are easy to satisfy by powering VCCIO from the FT232R's internal 3.3V regulator, via its 3.3V pin.
However there is still one other issue. If the design powers the FT232R from USB Vbus and the rest of it from another source (a very common design decision for low power battery gadgets), its essential *NOT* to drive the FT232R inputs when USB Vbus is not present. That's easy enough to do in software, just provide a potential divider that reduces Vbus to a convenient voltage for your MCU to sense, and if it goes away, immediately turn off the UART output and any related handshake outputs. In hardware its a bit trickier and the easy option may be to use level translating buffers rated for partial powerdown anyway, (e.g. 74LVC1T45 or 2T45) not for their level shifting but for their isolation when partially powered. A low threshold (~1.5V) N-MOSFET, gate to VCC/Vbus, and drain towards the MCU may also be suitable, if you put 10K from Vbus to Gnd to discharge its gate and the decoupling caps. That gives 1.7V 'headroom' to keep the MOSFET on when the MCU is outputting logic '1', assuming Vbus is at it nominal 5.0V. If Vbus is at the minimum permitted, it may clip the logic '1' level seen by the FT232R slightly. Note that this approach is impractical for isolating 5V logic.