The node between Q2's and Q6's collector puts out a current that is proportional to the input differential voltage; that is, the differential stage is a so-called transconductance amplifier. It doesn't produce a usable voltage on its own. To get a voltage from this, you'd typically use a PNP miller integerator (the one in the linked picture uses an NPN transistor, just flip it around for PNP). This integrator circuit will then put out a voltage that you can buffer with another NPN emitter follower. By choosing the integrator's capacitor appropriately, you can also stabilize the OpAmp (because it won't be stable if the gain doesn't roll off at high frequencies). This of course means that you'll have a very high gain at DC but the gain will drop off as the frequency increases. Building OpAmps with a constant high gain for all frequencies is sadly not physically possible, which is why at least one capacitor is required in every OpAmp - even in OpAmp chips. An unstable OpAmp will oscillate and not produce any usable output signal.
Here is an article about a very simple discrete OpAmp. You can clearly see the miller integrator formed by Q3 and C1. This circuit is the most basic OpAmp you can build and it's the basis for more complex circuits. You can improve it by adding current mirror loading to the differential pair, a current load for the miller transistor, darlington transistors at various places in the circuit and much more.
If you need more gain than this simple configuration can provide, you might need to look into cascode differential amplifiers, Wilson current mirrors and/or nested feedback loops. You might also need higher-bandwidth transistors so you can get more gain without your OpAmp becoming unstable. You sadly can't just add another common-emitter amplifier stage at the output of the OpAmp because that'll mess with the DC bias and also make the OpAmp unstable due to too much gain at high frequencies.
The output of the miller integrator in any OpAmp will have to be buffered, for example with an NPN emitter follower, because any loading on the miller integrator's output node will drastically reduce your overall open-loop gain.
Finally, in order to properly measure a discrete OpAmp, whether in reality or in a simulation, you'll need to provide external feedback to it so its operating point is properly stabilized. This will of course drop the gain of the complete circuit dramatically. To still measure the open-loop gain of the OpAmp alone in a simulation, you can apply an AC signal to the circuit and do an AC sweep. The expression "V(out)/V(in+,in-)" will show you the gain of the OpAmp itself (over frequency), where V(in+,in-) is the voltage difference between the OpAmp's non-inverting and inverting inputs. Alternatively, you can calculate the OpAmp's gain by multiplying the differential stage's transconductance with the miller integrator's DC transresistance.
.png){kind=link}