The uBitx is prone to overload if there are strong signals. I have done some tests with an analog AGC detector and different FET shunts on the RX line after the relay before the first mixer.
I used a similar detector circuit as other people have published here. However, I used a BC548 as an audio amplifier instead as this is a very common bipolar transistor that is highly available with a high Hfe. The Hfe is approx 200x on the B series . The gain was set to approx 200 in a common emitter amplifier. I take the signal from before the volume control. (Be aware that the volume control pot affects the impedance a bit so the level changes even if you tap on the hot side of the audio pot ... ). The detector is two 1N4148 diodes. There is a 1uF capacitor and a 1Meg resistor to set the first order RC circuit constant.
I tried with different FETs I had laying on the shack. Since this fet is used as a is a RF shunt circuit, it is not possible to use a JFET directly since JFETs are depletion mode.
By using an enhancement mode N channel Mosfet the RF shunt presents a high impedance to the AGC amplifier / recitifier. However, the output capacitance of the RF shunt Mosfet is a bit critial. I tried with a FQP30N06L (output capacitance in the 300pF range) and a IRLIZ44N (output capacitance in the 400pF range). Those presents a too low Z to the RF path since XC = 1/(2*pi*f*C). This is around 28ohm at 14 MHz, which will cause attenuation when the AGC is not active.
Instead I used a BSS138L N mosfet that I took of a level converter board I had laying around (I cut the board in four with the SMD part on to ease soldering). I soldered the mosfet controlled RF shunt directly to the relay pin and to ground on the back side of the uBitx circuit board instead of going in on the top side of the board as some other have reported on this forum. See pic.
Shunt mosfet for RF AGC:
The AGC seems to work well even with just one shunt element.
Testing with and without AGC: