HPGs are precious: they are capable of interstellar communication radically faster than anything else, and they are super expensive, costing at least 1B in TacOps.
Given the high price, it's economically critical to use a low cost multiplier vehicle. Given the enormous infrastructural value of the system, it's critical that it not be easily damaged. A Jumpship (x1.25) is easily damaged. A warship (x2) is somewhat tougher, but still easily tracked and damaged. Buildings stay in one place, so their location can always be traced with time, and then they are easily destroyed. The best approach I've been able to come up with is a support vehicle base submarine, along the following lines:
Oracle
3500 ton large naval vessel
Clan tech base
Cruise 9
Flank 14
Structure 32
Armor (BAR 10)
Turret 36
All others 38
Chassis: F, Submersible, Armored 1179
Engine: F, Fusion 1339
8 Lifeboats 56
Quarters (32+4 free) 320
Heat sinks 74
Armor (264 points) 14
Mobile HPG 50
Communications Equipment 15
2 LAMS (turret) 2
2 ERLL (turret) 8
2 LRM-20+AV (turret) 13
Ammo (24) 2
Turret 3
Fire Control (Advanced) 3
Active Probe 1
Angel ECM 2
Cargo 419
The cost is 1110M. (That's a system cost only 11% higher than the cost of the HPG itself.)
BV: 3546
The Oracle is that it's placed in an ocean where it runs deep for several years. The cargo includes 4 years of food and 4% spare parts and the luxury quarters for all should make life more viable. Oceans are probably pretty common---in the Sol system, aside from Earth, moons like Europa plausibly have a suitable ocean beneath a not-too-thick ice cover.
Most of the time, the Oracle simply handles a star system's HPG communications, including relays from elsewhere. This mission is furthered with excellent communication gear.
The ship has several nice capabilities which greatly enhance it's survivability in case someone wants to damage it or take it over.
The Oracle has a moderately anonymous location. This is very easy for incoming conventional communication, because that can be broadcast over a large area. It's harder for outgoing communication, but still feasible with various minor repeater station strategies.
Anonymity is pretty tough in general with HPG communications because they have a weak emergence wave. Nevertheless, a kilometer of water should greatly dampen any emergence wave (the Oracle can dive to 1980 meters in 1g before crush checks become nontrivial). For HPG reception, which has a much weaker signature, anonymity can plausibly be preserved.
Even when discovered by an enemy, it's difficult to damage. Nothing except another large submarine can reach similar depths, and the Oracle's 9/14 move can outpace almost all submarines.
Even when an enemy sub manages to close, the 2 LRTs (with Artemis V) are a very credible offense. And, even when the enemy manages to fire back, the Angel ECM and LAMS have a chance of zeroing out an incoming torpedo attack.
The ERLL are not really intended for combat: instead they are intended for burrowing. When encountering significant quantities of ice, the ERLL can break up and melt the ice in the vicinity of the Oracle, allowing it to slowly burrow. This use implies that an Oracle could also embed itself 10km beneath the surface of an iceball moon, where it would take more than a typical nuke to damage it.