Manamancer opens officially today. What it is, the real numbers behind it on launch day, the one thing its AI is not allowed to do, and what is still missing.
Manamancer is officially open as of today, September 21, 2026.
Here is the honest headline: almost nothing changed overnight. The site has been running in the open since the summer, and today it ships as version 1.0.0 — the 72nd tagged release. There is no big red button behind a launch like this one. What changes today is that we stop calling it a preview and start being answerable for it — for the decks you save here, the prices we show you, and the cards the AI recommends.
This piece covers four things: what Manamancer actually is, the real numbers behind it on launch day, the one thing its AI is not allowed to do, and — because a launch post that lists only strengths is an advert — what is still missing. There is a poll at the end about that last part.
What it is
Manamancer is a deck builder for Magic: The Gathering, covering Commander and thirteen other formats: Standard, Pioneer, Modern, Legacy, Vintage, Pauper, Alchemy, Historic, Timeless, both Brawls, Pauper Commander, and Canadian Highlander. Each format is a set of declared rules rather than a label, so a fifth copy of a card, a banned card, or a Sideboard that is one card too large shows up as a specific line naming the problem, not as a red badge you have to decode.
Around the builder sits the rest of what a deck needs. You can paste a decklist or a deck URL — thirteen importers cover fourteen sites, from Archidekt and MTGGoldfish to TappedOut and AetherHub. The stats panel draws your curve, your colour pips and your ramp, draw, removal and land counts against deckbuilding targets. A collection records what you own; a wish list records what you do not, works out what your decks are still short of, and splits a buy list across the sellers you chose. There is a goldfish playtest table, a catalogue of preconstructed decks, and an Explore page of decks other people have published. The interface ships in English, French, German, Spanish and Brazilian Portuguese.
Search, Explore, profiles, shared decks and the precon catalogue all work without an account. Saving decks, the collection, the wish list and the AI need one, and the account is free.
The numbers on launch day
We ask every article on this site to date its numbers, so the launch post does not get a pass. These were read from our own databases on September 19, 2026; the card index they describe was last rebuilt on September 18.
- Cards indexed
- 34,755 — 118,239 printings across 1,050 sets
- Daily price rows
- 83,037,676 — 102,231 printings tracked
- Price history from
- Feb 8, 2024 — One row per printing per day since
- Preconstructed products
- 925 — In 22 product lines
- Formats
- 14 — Each with its own rules engine entry
- Sellers compared
- 5 — Pick three, mark one as primary
Read from Manamancer’s card index, price archive and precon catalogue on September 19, 2026.
The card count is the least interesting number there and the most important design decision. Those 34,755 cards live in an index on our own server, rebuilt from Scryfall's bulk data every day at 09:00 UTC, rather than being fetched from someone else's API each time you type. That is why search never waits on a third party, why format legality is checked against the same data the search uses, and — this matters in the next section — why the AI can be held to a list of cards that provably exist.
The price archive is the number we are proudest of. A price on its own is trivia; 83 million of them are a history. Every paper printing we track has a row for every day back to February 8, 2024, which is what lets an article on this site say what a reprint did to a card instead of guessing what it will do, and what lets a card page show you the line rather than the dot.
The one thing the AI is not allowed to do
Every Commander deck has an Agent tab: an AI assistant that can see the whole list, reads the same curve and quota numbers the stats panel shows you, works out what the deck is trying to do, and recommends adds and cuts. Plenty of tools do something like that. The part worth explaining is the constraint.
The agent is not allowed to invent a card. It can only recommend cards that came back from a search of our index, and before any recommendation reaches your screen the server checks it again, three ways: the card exists, it is legal in Commander, and its colour identity fits inside your commander's. Whatever fails is dropped without ceremony. Ask it to fix the card draw in a mono-green deck and it cannot hand you Rhystic Study, however fond the internet is of it — blue is outside the identity, so the suggestion dies on the server and the chat never renders it.
What survives arrives as Apply and Dismiss chips beside the reasoning. Nothing in your deck changes until you press one.
Two caveats, because the guarantee is narrower than it sounds. First, validation proves a card is real and legal, not that it is right: the agent can still recommend a perfectly legal card that your deck does not want, and the Dismiss chip exists for exactly that. Second, it is Commander-only today. The mechanism is written up in full in the guide to the validated agent.
The AI is metered rather than gated. Every signed-in account gets a monthly allowance, free accounts included; the Pro and Max plans buy a larger one. What runs out is the budget, never the permission.
What is still missing
Three gaps, as of launch day.
First, the AI stops at Commander. The agent, primer generation and the full-deck generator all refuse other formats. A Modern or Pauper deck gets the builder, the legality checks, the stats and the prices, and no assistant.
Second, Moxfield links do not import. Moxfield puts its API behind a bot check that blocks automated fetching, ours included. Export the list there and paste it, and everything else — tags, Sideboard, Maybeboard — comes across. It is one extra step, and we would rather it were none.
Third, Cube and Two-Headed Giant are not formats here. Fourteen formats is a long list. It is not every way people play.
None of these is a secret, and we would rather you read them here than discover them on your second evening with the site.
Which gap should Manamancer close first?
- The AI agent in formats beyond Commander
- Cube support
- More sellers in the price comparison
- More interface languages
If you want to see what the site does before giving it an email address, start with Explore or the precon catalogue; if you would rather understand the numbers first, the guide to reading deck stats is the short version. When you are ready, make an account, import the deck you already have, and ask the agent what it would cut. Disagreeing with the answer is a supported workflow.
We will keep writing here — set reviews with the real prices next to the hype — and the latest releases are always listed in the app under What's New.
May your first deck here be the one your pod complains about, and may every card in it actually exist.
