Homebrew weapons and spells that won't break your table

Every homebrew item is a small contract with your table: this thing will feel special without making anyone else's turn worse. Most bad homebrew breaks the contract the same way, by adding power without subtracting anything. The official weapon list is boring precisely because it is balanced; every die size and property on it was paid for somewhere. Good homebrew keeps paying.

The encouraging news is that a weapon or spell needs far fewer decisions than people expect. Strip away flavour text and a 5e weapon is six or seven fields; a spell is a level, a school, an effect, and some delivery details. If you can fill in those fields honestly, the flavour can be as wild as you like. A sword that whispers the names of everyone it has killed is wonderful. A sword that deals 3d12 is a problem.

A weapon needs: a name, whether it is melee or ranged, which ability it attacks with (Strength or Dexterity), a damage die, a damage type, whether it is one-handed, and whether it counts as a simple weapon (which decides who is proficient with it). Optionally it can be finesse, carry a short/long range like 30/120, and have a special effect described in text. That is the whole skeleton; it is exactly the set of fields Loreweaver's homebrew form asks for, and it is enough to describe every weapon in the official game.

Balance lives in the relationship between the damage die and everything else. The official curve is roughly: a one-handed simple weapon sits at d6, a one-handed martial weapon at d8, a versatile or two-handed weapon buys up to d10 or d12 by costing you a hand or a shield. Every property you add should push the die down a step or hold it steady, never up. Finesse is the classic example: it lets the wielder use the better of Strength or Dexterity (which is exactly how the engine resolves it), and the official weapons pay for that flexibility by capping finesse damage at d8.

Damage type is the most underrated dial. The thirteen types (bludgeoning, piercing, and slashing, plus the ten elemental and magical types from acid to thunder) are mechanically identical until resistance shows up, and then they matter enormously. A longsword that deals radiant damage is numerically identical to a longsword, right up until the party meets something that fears the light. That is a lovely, nearly free way to make a weapon feel magical.

• One-handed simple: d6. One-handed martial: d8. Two hands buy d10 or d12.
• Finesse (best of STR or DEX) is worth about one die step; cap it at d8.
• An unusual damage type adds flavour and situational power at almost no balance cost.

A spell needs a level from 0 (cantrip) to 9, a school, an effect, and at least one class list it belongs to; casting time, range, components, concentration, and duration fill in the delivery. The class list decision is quietly important and usually skipped: a spell on the wizard list lives in a world of hundreds of alternatives, while the same spell on the ranger list might be the best option at its level by default. Put your spell only on the lists whose themes it genuinely fits.

Spell levels are power bands, and the cleanest calibration method is comparison shopping. Find the two or three official spells of the same level that do something adjacent, and ask whether a caster with one slot would ever be torn. If your 2nd-level blast simply outdamages every official 2nd-level blast with no strings attached, it is a 3rd-level spell wearing the wrong badge. Torn is the target: a real decision between yours and the classics means you have priced it correctly.

Two levers do most of the balancing work once the numbers are close. Concentration is a real cost: a concentration spell competes with every other concentration spell the caster knows, and only one can run at a time. And the save-versus-attack-roll choice changes texture more than power: attack rolls miss completely but crit; saves usually do something even on a success. When a design feels slightly too strong, adding concentration is almost always a fairer fix than shrinking the dice.

A few guardrails from Loreweaver's authoring model are worth adopting even if you homebrew entirely on paper. Damage entries are dice expressions, never flat numbers, because flat damage removes the variance that makes attack rolls exciting. Effect text is capped at 500 characters, which sounds like a spam limit but doubles as a design discipline: if a weapon's special rule cannot be stated in a short paragraph, it will be misremembered at the table, and a rule that is misremembered is not a rule. And homebrew cannot reuse the name of an official item; a player who hears "longsword" is entitled to assume it means what it has always meant.

Armour, if you go there, has the least room for creativity: a base AC, whether Dexterity adds to it (and any cap on that), and a weight category. The official range runs from 11 to 18 base, and every point above the curve for its category is a point of permanent, always-on power. Treat +1 over the official equivalent as a big deal, because it is.

Write your homebrew down in a structured form, not just prose in a campaign doc. Structured content can travel: Loreweaver exports a library as a small JSON file and imports it back, and it also reads 5etools homebrew files directly, converting their spells and weapons into its own format (long descriptions get trimmed to fit, and item types it cannot represent, like wondrous items, are reported rather than silently dropped). Whatever tools your table uses, the principle is the same: homebrew that exists in one player's head dies with the campaign; homebrew in a file outlives it.

Finally, playtest with permission to nerf. Tell the player up front: this is version one, and if it warps the game we will adjust it together. Most homebrew problems are not design failures, they are tuning passes that never happened. A d10 that becomes a d8 after two sessions is not a broken promise. It is the contract working.