Commander Deck Generator gallery

Krenko, Mob Boss

Requested: $50 mainboard budget, bracket 2, intent: none (commander's own plan).

Mainboard price
$49.98
Estimated bracket
3
Model calls
2
Tokens (in / out)
6,367 / 4,574
Estimated model cost
$0.0011
Wall time
89 s

What the generator itself flagged

These warnings are stored with the deck. They are shown unedited, including failures.

Commander profile

Written by the model in one profile_synthesis call from the commander's card text and the evidence below. Identity, set and timestamps come from trusted inputs, not the model.

Goblin tribal token swarm

Krenko's activated ability creates a number of Goblin tokens equal to the number of Goblins you control. This creates an exponential growth engine: each activation doubles (or more) your Goblin count, which in turn makes the next activation even larger. The deck's core strategy is to maximize the number of Goblins on the battlefield, protect Krenko so he can activate repeatedly, and then convert that massive token army into a win through combat damage, direct damage (Impact Tremors, Goblin Bombardment), or alternative win conditions. The deck follows standard card evaluation for tribal commanders: it wants the best Goblin creatures and token producers, with a heavy emphasis on untapping Krenko and reducing his activation cost.

Win conditions the profile proposes

"in deck" marks profile key cards that made the final list. Others were cut by budget, legality, or optimization.

Build paths the profile describes
  • Classic Goblin Tribal (mainstream): The mainstream build: maximize Goblin count with the best tribal support cards, lords, and token producers. Focus on curving out with Goblins, protecting Krenko, and overwhelming opponents with sheer numbers.
  • Untap Combo (emerging): An emerging build that uses untap effects on Krenko to generate exponential or infinite tokens in a single turn, then wins through a damage outlet or combat.
  • Goblin Storm (underexplored): A text-supported but underexplored path that uses the massive token count as a mana source (Skirk Prospector, Brightstone Ritual) to cast big spells or chain Goblin spells, potentially looping with cards like Goblin Recruiter and Conspicuous Snoop.
Rest of the strategic profile
Synergy priorities
high: ['Goblin creature count', 'Untap effects for Krenko', 'Token production', 'Haste enablers', 'Protection for Krenko']; medium: ['Goblin lords', 'Damage outlets for tokens', 'Card draw from tokens (Skullclamp)', 'Mana generation from Goblins']; low: ['Non-Goblin token producers', 'Generic red burn', 'Board wipes']
Anti synergies
description: Cards that remove Goblins from the battlefield or prevent token creation, cards_to_avoid: ['Damping Matrix', 'Cursed Totem', 'Linvala, Keeper of Silence'], reasoning: Krenko's ability is an activated ability that requires tapping. Effects that prevent activated abilities from being activated completely shut down the deck's primary engine.; description: Board wipes that reset the Goblin count, cards_to_avoid: ['Blasphemous Act', 'Anger of the Gods'], reasoning: The deck's power scales with the number of Goblins on the battlefield. A board wipe resets the engine to zero, forcing a slow rebuild. While some decks can recover, Krenko decks are particularly vulnerable because they rely on a critical mass of creatures.; description: Cards that replace Goblins with non-Goblin tokens, cards_to_avoid: ['Myr Battlesphere', 'Pentavus'], reasoning: These cards produce tokens that don't count as Goblins, so they don't feed Krenko's engine. They may look like token producers, but they don't advance the core strategy of increasing the Goblin count.
Strategic constraints
mana_base_requirements: Mono-red, so the mana base is simple. Basic Mountains are sufficient, though utility lands that untap creatures or produce extra mana can help. The deck wants to cast Krenko on turn 3-4 and then activate him repeatedly, so mana acceleration is important.; interaction_density: low; speed_tier: fast
Power indicators
estimated_ceiling_bracket: 4; estimated_floor_bracket: 2; notes: Krenko can be built as a fast, aggressive deck that wins quickly with a critical mass of tokens. With untap combos, it can win on the spot. However, it is vulnerable to board wipes and interaction that removes Krenko. The floor is a casual tribal deck; the ceiling is a high-powered combo deck.
Value inversions
Engine dependencies
engine: Goblin creature count on the battlefield, engine_card_traits: ['Creature with type Goblin', 'Token creation that makes Goblin tokens', 'Effects that copy or double token creation'], dependent_outputs: ['Massive token army for combat', 'Mana generation via Skirk Prospector or Brightstone Ritual', 'Damage via Impact Tremors or Goblin Bombardment', 'Card draw via Skullclamp'], false_synergy_warning: Cards that produce non-Goblin tokens (e.g., Myr Battlesphere) or that deal damage without involving Goblins (e.g., direct burn spells) look like they fit the token/damage theme but don't increase the Goblin count, so they don't feed the engine. They are traps that dilute the deck's focus.
Mispriced card examples

Evidence cited

SourceWhat the build had
Card textOracle text for Krenko, Mob Boss: {T}: Create X 1/1 red Goblin creature tokens, where X is the number of Goblins you control.
Commander cohort (EDHREC)available: 44,314 public decks, general cohort, retrieved 2026-10-02 from EDHREC. observed inclusion rates only; no win rate or causal effect is implied; edhrec general cohort, 44314 decks, 220 cards
Most-included cards given to the profileMountain 98% in deck, Goblin Warchief 87% in deck, Skirk Prospector 86% in deck, Sol Ring 85% in deck, Impact Tremors 85% in deck, Goblin Matron 84% in deck, Goblin Chieftain 77% in deck, Goblin Bombardment 74% in deck, Brightstone Ritual 68%, Pashalik Mons 68% in deck, Siege-Gang Commander 65% in deck, Goblin King 64% in deck, Krenko, Tin Street Kingpin 63% in deck, Hobgoblin Bandit Lord 62% in deck, Skullclamp 62% …
Community discussion (Reddit)0 threads. The search request was refused (HTTP 403) during these builds, so the profile had no community evidence.
Rules and reference chunks0 referenced. The reference corpus in the build database is empty, so no rules text was retrieved.
Tournament results0 decks; no tournament signal for this build.
Scoring signalsUsed: commander_cohort, embeddings, narrative, rules. Unavailable: edhrec, tournament_cwe. The legacy edhrec signal (a per-card table filled by a separate ingestion job) is reported unavailable even though the cohort data above was used; the project's evaluation report lists merging these two labels as follow-up work.

Pipeline stages

Deterministic code does the assembly. Counts and timings below are what this run recorded.

  1. Hard filters

    Color identity, Commander legality, singleton and per-card price ceiling leave 7,132 candidate cards. (1.0 s)

  2. Role scoring and Pareto pruning

    Each candidate is scored on synergy, mana efficiency and replacement value. Candidates dominated on every objective are dropped, leaving 941. 16 protected strategy and infrastructure cards are reserved before pruning so they cannot be lost. (1.4 s)

  3. Strategy plan

    Archetype typal, status satisfied. Requirements: Creature type: goblin ≥ 16. Final counts: Creature type: goblin 30.

    Requirements verify supported engine ingredients, not a complete win or competitive strength.

  4. Role targets

    Targets come from the commander template, which bounds what the optimizer aims for.

    RoleTargetFinalStatus
    Ramp1110short
    Card draw76short
    Removal45met
    Board wipes20short
    Win conditions33met
    Tutors25met
    Protection30short
    Recursion21short

  5. Infrastructure and synergy fill

    Infrastructure placed first: Sol Ring, Arcane Signet. A greedy fill then adds the card with the best marginal objective until 99 slots are filled (synergy matrix over 941 candidates).

  6. Swap refinement

    The optimizer proposed 16 one-for-one swaps, each improving its current objective. Under the production preserve policy, a change set must pass a function-preservation check. This one was rejected (supported_function_loss, unknown_removed_functionality, missing_one_to_one_substitution_proof, deck_support_loss_or_unmodeled_dependency), so the previous list was kept.

  7. Budget repair and rebalance

    The optimizer proposed 1 upgrade(s), 1 unbundle(s) and 0 downgrade(s), for a mainboard total of $49.61. Under the production preserve policy, a change set must pass a function-preservation check. This one was rejected (supported_function_loss, unknown_removed_functionality, missing_one_to_one_substitution_proof, deck_support_loss_or_unmodeled_dependency), so the previous list was kept.

    Mainboard total after this stage: $49.26 (98.5% of budget).

How the list changed between stages

SnapshotCardsMainboard priceChange from previous
Synergy fill (greedy assembly)99$49.26Starting list
After swap refinement and strategy-variant choice99$49.26No change
After budget rebalance and repair99$49.26No change
After the safety review's replacements99$49.26No change
Final list99$49.98
1 in, 1 out

In: Lightning Bolt

Out: Shock

Optimizer 4.8 s, review 34.6 s, profile 46.2 s.

Safety review

One batched card_fit_batch call reviews the riskiest selections and a menu of alternatives. Bounded risk review, not exhaustive card review.

Reviewed 18 of 47 eligible selections, plus 8 alternatives, in 1 call. A replacement must pass its own verdict and the final legality, budget and role checks. If no replacement passes, the card stays in the deck and is flagged, as below.

Selections left flagged

Flagged cardWhy it stayed
Mogg FlunkiesNo reviewed replacement satisfies final constraints.
Akoum StonewakerNo reviewed replacement satisfies final constraints.
Self-AssemblerNo reviewed replacement satisfies final constraints.
Tormenting VoiceNo reviewed replacement satisfies final constraints.
Merchant of the Vale // HaggleNo reviewed replacement satisfies final constraints.
Ardent ElectromancerNo reviewed replacement satisfies final constraints.
Valakut PredatorNo reviewed replacement satisfies final constraints.
ShockNo reviewed replacement satisfies final constraints.
Grotag Siege-RunnerNo reviewed replacement satisfies final constraints.
Infectious BloodlustNo reviewed replacement satisfies final constraints.
Skirk CommandoNo reviewed replacement satisfies final constraints.
Rock JockeyNo reviewed replacement satisfies final constraints.
Welder AutomatonNo reviewed replacement satisfies final constraints.
Coal StokerNo reviewed replacement satisfies final constraints.
Molten FrameNo reviewed replacement satisfies final constraints.
Shredded SailsNo reviewed replacement satisfies final constraints.
Goblin Rock SledNo reviewed replacement satisfies final constraints.
14 deterministic capability findings on cards in this deck

Final legality check

Recomputed for this page from the stored card records, separately from the builder's own acceptance check.

CheckResultDetail
Commander is legal as a commanderpassKrenko, Mob Boss
99 cards plus the commanderpass99 + 1
Singleton (basic lands and 'any number' cards exempt)passno duplicates
Every card within the commander's color identitypassidentity R
Every card Commander-legal in the stored catalogpassall legal
Mainboard within budget (commander priced separately)pass$49.98 of $50; commander $2.92

The builder's own acceptance check recorded 0 hard failure(s) and 9 warning(s), listed at the top of this page. Prices are from the corpus snapshot, not today's market.

Tokens and cost

Every model call goes through one client that records token usage, linked here to this deck. Costs are estimates at the configured rates ($0.06/M input, $0.015/M cached input, $0.18/M output for deepseek-ai/DeepSeek-V4-Flash-0731:deepinfra), not invoices. The deck summary is rendered from card facts and makes no model call.

Model callInput tokensCached inputOutput tokensEstimated cost
profile_synthesis3,1212,8162,692$0.000545
card_fit_batch3,2465121,882$0.000510
Total (2 calls)6,3673,3284,574$0.001056

Decklist

Grouped by the slot role the optimizer assigned. ⚑ marks a card the safety review left flagged.

Ramp 10

Card draw 4

Removal 5

Win conditions 3

Synergy and utility 43

Lands 34

Download the JSON record for this build, which includes the full profile, the stage snapshots, review and audit data, and model-call usage.