We report new experimental data on the composition of magmatic amphiboles synthesised from a variety of granite (sensu lato) bulk compositions at near-solidus temperatures and pressures of 0.8–10 kbar. The total aluminium content (Al$\mathrm{tot}$) of the synthetic calcic amphiboles varies systematically with pressure ($P$), although the relationship is nonlinear at low pressures (<2.5 kbar). At higher pressures, the relationship resembles that of other experimental studies, which suggests of a general relationship between Al$\mathrm{tot}$ and P that is relatively insensitive to bulk composition. We have developed a new Al-in-hornblende geobarometer that is applicable to granitic rocks with the low-variance mineral assemblage: amphibole + plagioclase (An$\mathrm{<mrow\; data-mjx-texclass="ORD">15<mrow\; data-mjx-texclass="ORD">–</mrow>80</mrow>}$) + biotite + quartz + alkali feldspar + ilmenite/ titanite + magnetite + apatite. Amphibole analyses should be taken from the rims of grains, in contact with plagioclase and in apparent textural equilibrium with the rest of the mineral assemblage at temperatures close to the haplogranite solidus (725 ± 75 °C), as determined from amphibole–plagioclase thermometry. Mean amphibole rim compositions that meet these criteria can then be used to calculate $P$ (in kbar) from Al$\mathrm{tot}$ (in atoms per formula unit, apfu) according to the expression:

$P$ (kbar) = 0.5 + 0.331(8) × Al$\mathrm{tot}$ + 0.995(4) × (Al$\mathrm{tot}$)$2$

This expression recovers equilibration pressures of our calibrant dataset, comprising both new and published experimental and natural data, to within ±16 % relative uncertainty. An uncertainty of 10 % relative for a typical Al$\mathrm{tot}$ value of 1.5 apfu translates to an uncertainty in pressure estimate of 0.5 kbar, or 15 % relative. Thus the accuracy of the barometer expression is comparable to the precision with which near-solidus amphibole rim composition can be characterised.