Here's list of Mage cards that aren't in need of changes at all:
Commons: Wizard, Acolyte, Elven sorceress.
Uncommons: Elven enchantress, Red mage, Sorceress, Summon master, Warpriest.
Rares: Elven warlock, Lord of time, Luna, Myth summoner, Thor, Wind witch.
Chaos wizard costs 4G and discards a common from hand. No additional effect, nothing more on top of that. That's definitely not enough. How about something like this: if a card was discarded, add 3 tower, 5 wall or 2 stock?
Conjurer has some potential as resource card, but production x0 is too prohibitive, especially when you have Mage/Legend deck. I'd rework it like this:
Replace a card in hand with Conjurer
If last card played was Conjurer
Druid's summoning effect has little to no synergy with Mage, mostly because you don't want, say, Deathwood shaman in non-Nature and non-Undead deck. But I think it doesn't need changes.
Fire guardian - one of the few defensive cards for Mage deck. But for Mage deck it's pretty expensive - 6g. I'm not sure if Mage deck needs another common Mage/Burning or it needs "Mage" in its condition.
Librarian - Mr. Tokens is crucial for any Mage deck, but it doesn't make it even particularly good. Remove its attack and add Quick.
Warlock - this card doesn't worth to play it in 99% cases. Deadweight for Ice queen or replacing cards.
Apostate - just remove Mage keyword from it, it would become effective only with something like "Magic: +1, Mage tokens: +40", which won't (and, what's more important, shouldn't) happen at all.
Elementalist. Swift, Banish, Mage - good keyword combination, but completely unneeded Unliving bane. Maybe add Aria and remove Banish or raise its cost to 15-16?
Knight of time. With, say, 10G/9R cost it would be much better.
Archmage is bad, but I can't say anything on top of my posts in card discussion.
Master of the past. Really, I'm not sure if it must be changed al all... but it doesn't counter any stock- cards like Tornado. So I dare to say it could use 0-cost.
Now let's look at support cards.
List of cards that are fine:
Commons: Mage guild, Fireball, Arcane bolt, Devilkin.
Uncommons: Magical clay, Magical steel, Meditation, Valley of thousand rivers.
Rares: Shadow unicorn.
Wizard's cottage: cheaper and weaker Mage summoner than Mage guild. I'd add Gems production x2 and raise its cost to 5B/1G, similar to Orc banner.
Stonehenge: compare it with Centaur trainer. Pretty much the same cost, but Centaur trainer has two powerful keywords (well, Beast isn't so powerful, but rare Beasts can be summoned quite easily), while Stonehenge doesn't have anything on top of Magic+ and summoning. Moreover, you need bricks to play it, and it means that you won't be able to play something defensive like Fortification. I see two ways to fix it: lowering its cost to, say, 13B/12G or adding some tower/wall gain (say, 7 tower/10 wall).
And last, but not least - Ice queen. Don't get me wrong, I really like it. But if you don't have it in hand, you're in a world of crap. I suggested to make it more reliable several times: Glacial blessing concept, change to Archmage, change to Myr's academy and suggestion about Mage keyword to Ice queen itself. None of those was implemented. Again, Ice queen should be more reliable, even with weaker effect.
And the final part: keyword. I've asked for higher token gain, and I think it would be a good decision. Another possible change - raise potential Magic gain from it to 3 for better synergy with Magic+ cards.
That's what I can suggest to give Mage decks some love. Any kind of feedback is appreciated, as usual.