Really functioning and
unambiguous (!) definitions of the terms direct/indirect/inverse/reverse/silent agonist resp. antagonists look a bit more complicated than the ones cited so far in this thread, no offense intended though. I think I could write a short summary soon to explain what I mean, because this kind of confusion seems to appear over and over again. In short just one note: All these definitions changed throughout the last decades several times and the current version(s) strongly depend on the most recent results in the protein-related fields. There are no simple black'n'white patterns possible as definition, as the vast majority of effector proteins in nature do
not work in a binary fashion (ON/OFF) but by graduaded, tunable, modifiable ways.
Anyway, the point is that
tolerance is theoretically possible for all kinds of physiologically active ligands:
1st: It doesn't matter if a ligand binds to its target effector and simply inhibits it to perform its duty, may it be to intracellularly release a second messenger, to open a ion-, water-, whatever-selective channel/pore, to catalyse a chemical reaction, to change its conformation, to break up into two or more fragments resp. the contrary, to aggregate to dimers, trimers, ...
2nd: It doesn't matter if a ligand increases the activity of said duty.
3rd It doesn't matter if a ligand decreases this activity, either to a lower level or entirely to zero (...the latter case would equal the "1st" point above, ie. nothing at all is happening). As a side-note: Per definition there can't be negative values for a protein's activity. Either a protein does something or it does nothing and is simply floating around silently resp. sticks lazy to a membrane. What is actually possible, is that it does something else. Many proteins are able to perform more than one task, resp. many (GPCR-)receptors can cause more than just one reaction (see next point).
4rd: It doesn't matter if ligand B (eg. a drug) causes a protein to do something entirely different than the usual ligand A (eg. the endogenous neurotransmitter).
I hope I didn't forget any prominent possibility
...in any of the above cases development of tolerance
is possible: I would bet some money that one could find examples simply by searching through PubMed with the right search terms. Because any of the above cases will cause an interference with the normal neurochemistry (to stick with psychoactive drugs for now), there
has to be some kind of physiological reactions. Of course do we know exceptions (eg. [almost] no tolerance to miosis and obstipation from opioids), but then again are only extremely few regulatory systems in nature absolute.