fvtt-mcp-molten5e
This server is a D&D 5e MCP bridge that lets Claude Code read and edit a live Foundry VTT world on Molten Hosting — authoring actors, items, scenes, journals, and more, plus managing static assets over WebDAV.
Actors & PCs — list/get/search actors, copy premium-book creatures, author custom NPCs, build and level PCs with real dnd5e advancement, set ownership, delete actors.
Items & equipment — create/edit/delete world items, import official gear from compendiums, author structured weapons/armor/consumables/containers, manage activities (attacks, saves, heals, cast) and active effects/conditions.
Compendium discovery — search premium PHB/DMG/MM packs for creatures, spells, and items with faceted filters; fetch full stat blocks.
Journals & quests — build styled multi-page quest journals, link quests to real NPCs, append session logs, search, and control player visibility.
Scenes & placeables — create/update/delete scenes and place walls, lights, sounds, tiles, drawings, tokens, map-note pins, and regions/teleporters; screenshot scenes for visual QA.
Combat & world state — apply conditions, read combat stats and chat logs, manage the in-world calendar and dnd5e 6.0 automation settings.
Asset file management (WebDAV) — list, upload, download, move, copy, delete, and URL-map files under Data/, with reference-aware safety checks.
Reference integrity — find and relink asset references across documents.
Adventure map-pack import — read Tom-Cartos-style scene-pack modules off disk and recreate their scenes/journals in the world.
Provides tools to turn Discord voice recordings (via Craig) into session transcripts, recaps, combat reports, and GM notes aligned with the Foundry chat log.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@fvtt-mcp-molten5e list all NPCs in the current scene"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
fvtt-mcp-dnd5e
An MCP server and a set of Claude Code skills that build D&D 5e content in a live Foundry VTT world. A pasted stat block becomes a complete NPC with its loot; a map image becomes a walled, lit scene; an adventure becomes journals, tables and handouts; a Discord recording becomes the session recap. Everything is built from the 2024 premium books you own, never the SRD.
Targets dnd5e 6.0.5+ on Foundry 14.368+, on any host: a hosting provider, this machine, a URL.
Setup
git clone https://github.com/Txpple/fvtt-mcp-dnd5e && cd fvtt-mcp-dnd5e
npm install && npx playwright install chromium && npm run build
npm run init # asks, writes .env and the registration, runs the doctor
node scripts/install-skills.mjs ~/my-campaign # optional: link the skills into another projectnpm run init first says the one thing it cannot do: the Foundry user the server joins as must
exist in the world (Users → Create User; Gamemaster or Assistant GM, default name MCP-Claude).
Then it asks where Foundry runs (local, molten or generic), its URL, that user and its
password, and the admin password (the Setup screen's Administrator Password, not the license key),
plus the wake URL and File Manager password on Molten or the Data folder of a local install. It
writes those to .env (an existing one is backed up and changed in place), registers the server
for Claude Code with the full Node path (this clone's .mcp.json, another project's, or
~/.claude.json for every project; never a secret), runs the doctor as its join report, and
prints the same entry as an mcpServers block for any other MCP client. For a cautious first run,
answer yes to read-only (FOUNDRY_TOOLSETS=readonly) and join as a Player or Trusted Player: the
server then reads the world and refuses every write (Tools, below). Every answer is also a flag
(npm run init -- --help; --yes asks nothing). By hand instead: the minimal block at the top of
.env.example into .env, .mcp.json.example into .mcp.json with absolute paths, then
npm run doctor.
npm run doctor checks, in order: Node, the build, Playwright's Chromium, the .env (as the
server reads it), the Foundry app (local host), that Foundry answers, the admin key (when no world
is running), the world, a login as FOUNDRY_USER with its role, and the bridge's renderer and
Open Roll 5e modules. Every ✗ line names its fix, and any ✗ exits non-zero. It never launches or
stops a world; the login is the same short join a tool call makes, so it is safe against a live
instance.
Start Claude Code and say "start the world": the server wakes the instance if it sleeps,
launches the world if FOUNDRY_ADMIN_KEY is set, joins, and reports the world, the host, its role,
which premium books it found, and the bridge's state (Chromium's renderer, the Open Roll 5e modules
it found active, the toolsets it advertises, and a warning for anything that needs a look). A user
name the world does not have is refused at the first tool call, with the users it does have. Then
ask for content.
Requirements: Node.js 22+ · Foundry 14.368+ with dnd5e 6.0.5+ (6.0.0–6.0.3 delete every
effect without an expiry on a rest; the tools refuse a pre-6.0 world; an older world is upgraded
Foundry-first, then dnd5e:
docs/hosts.md) · the Monster Manual,
Player's Handbook and Dungeon Master's Guide modules for authoring (Heroes of Faerûn and
Ravenloft: The Horrors Within are optional).
Related MCP server: FoundryVTT MCP Server
Skills
Say it in plain words; the matching skill calls the tools.
Skill | Does |
| boot the world and report its state (read-only) |
| a complete NPC from a stat block: stats, actions, spells, effects, inventory, loot, art |
| a complete player character: class, species, background, choices, spells, gear, art, owner |
| weapons, armor, wondrous items, potions, loot; the real PHB / DMG item first |
| a map image into a ready-to-play scene: walls, lights, mood, playlist, journal |
| quests, handouts, lore, boxed text, GM notes, recaps, linked to each other |
| roll tables, card decks, music and ambience |
| atmospheric sound for a scene (companion module, below) |
| a scene-pack module into your world: scenes, walls, lights, teleporters, legend |
| a cut-out token image onto an actor |
| narration, NPC dialogue, whispers, roll requests, item cards; export or prune the log |
| audit next session's encounters; diff the world against the plot; log what the party fought |
Session summaries and analytics moved out in 4.0.0, to
fvtt-mcp-sessionscribe. That covers turning
a Craig (Discord) recording and the chat log into the transcript, recap, combat report and GM
notes. It files the session diary through manage-journals here.
The campaign-facing skills keep your campaign's facts and house style in a repo of your own
(campaign.json + STYLE.md: campaign-repo.md);
nothing about one table lives here.
Tools
80 tools. One family tool per document type, selected by action: manage-actors,
manage-scenes, manage-placeables (with a kind: walls, lights, tokens, regions, sounds,
tiles, drawings, notes), manage-items, manage-journals, manage-rolltables, manage-cards,
manage-playlists, manage-folders, manage-macros. Around them:
D&D authoring:
author-npc,create-pc/level-up-pc,add-feature,add-item,manage-effect,manage-activity,apply-condition,set-actor-art, the quest journals,content-audit.Reading:
get-world-info,search-compendium(creatures / spells / items, with facets),get-compendium-entry,read-pack,search-journals,screenshot-scene.At the table:
send-chat-message,request-roll,post-item-card,export-chat-log,roll-on-table,configure-combat-tracker,activate-scene,configure-dnd5e-settings,manage-calendar.Files: upload, list, download, copy, move, delete, relink; on every host through Foundry's own file picker, faster with a direct plane.
Users and organization: ownership, groups and the primary party,
move-documents,bulk-delete.Scripting:
run-script— a short program against the page functions behind every tool (fvtt.*, the reference): a loop or a bulk job in one round trip, with writes capped and recorded, a dry run, and undo by receipt. It runs with the bridge user's rights: that user's role, not the facade, is the boundary.
A list answers one line per record, a get answers JSON, a miss is an error, an unknown argument
is refused by name. Every write to the world's documents answers with a receipt, which
run-script { "undo": "rcpt_…" } reverses; a change outside them (a file in Data/, a player's
view) cannot be undone, and its answer ends with a "not undoable" line saying what it changed. The
full table is src/registry.ts. FOUNDRY_TOOLSETS
lets a registration advertise a subset (chat,combat is 11 tools instead of 80).
FOUNDRY_TOOLSETS=readonly advertises every toolset's reads (41 tools) and refuses every other
call by name before it runs; pair it with a Player or Trusted Player join user, whose role is the
real boundary (docs/hosts.md).
One tool needs a companion module and warns when it is absent: configure-area-sounds
(Open Roll 5e: Area Sounds).
Hosts
Where the world runs is a preset chosen per registration with FOUNDRY_HOST
(docs/hosts.md):
| Means |
| any Foundry at |
| Molten Hosting: the panel's wake URL, WebDAV derived from the URL |
| an install on this machine, its |
One .env serves every registration; .env.example explains each variable.
FOUNDRY_URL_FALLBACK names a second path to the same instance (a tunnel beside a VPN name), tried
once when FOUNDRY_URL times out on the way to game.ready; get-world-info's host block says
which one the session is on.
How it works
Foundry has no server-side plugin API, so the server drives a headless Chromium (Playwright):
it joins the world as the dedicated user, waits for game.ready, and injects a page-side library;
every tool call runs inside that live page through Foundry's own client APIs. The connection is
lazy: the first real tool call is what wakes and joins. The other Open Roll 5e servers and the
modules' live suites drive a world through the same bridge as a library:
import { connectFoundry } from 'fvtt-mcp-dnd5e/client' (docs/contracts.md).
While connected the bridge leaves a seat record (a file on the machine, a marker on its User
document) naming the session that holds it; disconnect-bridge names any other holder and ends
them with { "all": true }, and a bridge idle for 20 minutes logs itself out
(FOUNDRY_IDLE_LOGOUT_MIN) — docs/hosts.md. A companion's live test run
(By the Books' battery) takes a lock on the whole instance for its length; the bridge refuses to
join while one is held, naming the run and the suite it is on, and local-foundry.mjs status
shows the seat — docs/local-sandbox.md, The seat lock.
Development
The offline gate is npm run check && npm run typecheck && npm test && npm run build && npm run knip
(npm test prints and ratchets the context budgets). Live proof runs against a sandbox:
FOUNDRY_HOST=local node scripts/verify-<family>.mjs and
FOUNDRY_HOST=local RUN_LIVE=1 npm run test:integration. House rules:
CONTRIBUTING.md · release gate: docs/RELEASE.md · design:
design.md · changes: CHANGELOG.md.
Out of scope: other game systems; a live in-session assistant; AI map generation; D&D Beyond import.
Security
Outbound only: the server and the browser run on your machine and authenticate to Foundry as a
user. Secrets stay in .env (gitignored); errors name a missing variable, never its value. Agent
inputs are untrusted: writes go through Foundry's own APIs, destructive file operations refuse a
live world's database and delete only on an exact match. Anything under Foundry's Data/ is
served to the world's users, so don't upload anything sensitive.
Part of Open Roll 5e
fvtt-mcp-dnd5e is one of the three MCP servers in Open Roll 5e, a suite of Foundry VTT modules and Claude Code tooling built for one D&D 5e table and shared. The other servers:
fvtt-mcp-imagegen: makes the art with Google's Gemini image models: icons, tokens, props, portraits and illustrations, token redresses and restyles, battlemap and overland-map repaints, and the illustrated session records, all grounded in what the world already shows.
fvtt-mcp-sessionscribe: turns a session's Discord recording and Foundry chat log into its record. Its end-to-end
session-scribeskill drives the server from the Craig link to a speaker-labelled transcript, a fully illustrated player recap, combat statistics, GM notes and a party snapshot.
The modules, each of which installs and works on its own and none of which needs another:
Open Roll 5e: Autoexplore: lets a scene start fully explored, so the whole map shows through the fog of war while tokens still need line of sight.
Open Roll 5e: By the Books: the 2024 rules played from the sheet: a hit rolls and applies its own damage, saves resolve themselves, reactions hold, concentration is tracked, and rests, senses, sizes and lights follow the book. Every rule that rides a use or an effect in the 2024 core books, Heroes of Faerûn, Arcana Unleashed and Ravenloft: The Horrors Within, each step with a receipt and a revert; offered, never enforced.
Open Roll 5e: Combat Plus: automates the chores of running a fight: combat music, an initiative gate, an out-of-turn movement block, defeated marking at 0 HP and turn alerts.
Open Roll 5e: Errata: corrects, in memory, bugs in the premium D&D 2024 books, the dnd5e system and Foundry itself, each fix held until the vendor ships its own.
Open Roll 5e: FX Studio: visual and sound effects for dnd5e, played from what actually happened at the table, with about a thousand stock FX and a window for authoring your own.
Open Roll 5e: Loot Shelf: loot chests and merchant shelves that players can take from, buy from and sell to without owning them, with a receipt for every trade.
Open Roll 5e: Open Server: for hosted worlds: clears the startup pause so players can play before the GM arrives, and gives every user a one-click jump to their character's token, and their familiar's, from the Players list.
Open Roll 5e: Party Stash: makes a dnd5e Group actor's inventory a working party stash: drags move instead of copying, coin moves through a dialog, and every transfer posts a receipt.
Open Roll 5e: Area Sounds: background sound for scenes: random one-shots with silence between them, seamless crossfaded loops, day and night gating, and quiet during combat.
Open Roll 5e: Token Art: gives every token the look of its state, ready, in combat or dead, each a file named beside the base art, with nothing to configure.
Open Roll 5e: Clue Board: keeps the party's Clue Board: every clue the DM hands them, stamped with where it was found and who has read it, pinned to the threads the party names.
Open Roll 5e: Card Catalogue: turns every directory the table uses into a case of cards: the journal, the scenes, the actors, the items and the compendium packs, each found by any word with its main action a click away, and a sidebar that shows only the tabs the table uses.
Open Roll 5e: Crystal Ball: a crystal ball on the sidebar: a player asks a Claude model about the game as their character knows it, the journal they can read, their sheet, their party and its clues, answered by a service beside the server, never from a client.
Issues are welcome on every repo in the family; pull requests are not accepted, since each is one author's design for one table, shared because it might suit yours. How they fit together is mapped in fvtt-suite-openroll5e.
License
MIT. See LICENSE.
Available Tools
151 toolsactivate-sceneA
ACTIVATE a scene — make it the one active scene, which is where connected clients land and where users log in. Foundry allows exactly ONE active scene world-wide, so this deactivates whatever was active; the previous scene is reported back so you can put it back. Activating an already-active scene is a no-op and says so (alreadyActive) rather than pretending it did something. Kept separate from update-scene (which is scene-DOCUMENT-only and never activates) because this changes every connected player's screen. To move only SOME players, leave the active scene alone and use pull-users-to-scene instead. GM-only.
Inputs: sceneIdentifier (required).
add-featureA
Add a feature/spell/ability to an existing actor (NPC or PC). Set mode: • 'compendium-features' — import named class/monster features from an official compendium (PREFERRED for official content, e.g. Pack Tactics, Multiattack, Spellcasting). Params under compendiumFeatures. • 'feature' — author a feature/attack/spellcasting setup/spells from scratch (use only when not available in a compendium). Params under feature (select feature.featureType). • 'items' — attach world items by raw data. For real GEAR prefer import-item (copy from a compendium, keeps art+stats) or add-item (author); use this mode only for free-form item data. Params under items[]. actorIdentifier (exact name or ID) is always required — find it with list-actors / get-actor.
Inputs: mode (required), actorIdentifier (required), items, feature, +1 more.
add-free-castA
[D&D 5e] Grant "cast without a spell slot, N per rest" the native 2024 way — Magic Initiate, Favored Enemy, lineage grants. TWO sheet entries result: the spell stays in the repertoire as a normal ALWAYS-PREPARED spell (castable with slots; imported from the compendium if missing), and a cast activity ON the granting feature projects a " - " entry into the sheet's native "Additional Spells" spellbook section with its own tracked pool (default 1/long rest) — no slot, no use-dialog. Also MIGRATES the old shape (on-spell use pool + forward activity) off the spell, and dedupes dnd5e's cached spellbook copies. Idempotent. NEVER track a free cast as a separate tracker feat or as a forward on the spell.
Inputs: grantedBy (required), actorIdentifier (required), spellIdentifier (required), uses, +1 more.
add-itemA
[D&D 5e only] Create a structured physical item (loot/gear) on an actor or in the world Items sidebar. Pick itemType, then supply only the fields you need — sensible defaults fill the rest:
• weapon — to-hit weapon. damage (base die), weaponClass, attackType, reach/range, magicalBonus, properties. Builds a rollable attack activity by default (withAttack).
• armor / shield — armorValue, dex, strength; magicalBonus = +N AC. Pass wireAc (BODY ARMOR only) to make the actor derive AC from the worn armor; a shield needs no wireAc (its +2 applies under any AC calc).
• wondrous — rings/cloaks/etc. (equipmentType); use magical:true + attunement (a wondrous item has no numeric +N field — model a bonus with manage-effect).
• consumable — potion/scroll/ammo/wand. consumableType, uses {max, recovery, autoDestroy}. Ammo can carry damage + ammoReplace + magicalBonus.
• tool — toolType, ability, proficient, toolBonus.
• loot — gems/art/trade goods (lootType, price). NOT equippable/attunable.
• container — bag/chest with capacity and an inner currency pile. Place items inside any container with the container param (id or name).
Cross-cutting: price, weight, quantity, rarity, identified, equipped, attunement (""/required/optional) + attuned, magicalBonus (the +N), properties (incl. "mgc"). Setting magicalBonus/magical adds the mgc flag; the numeric +N is stored for weapons, body armor, and magic ammo (wondrous/potion have no +N field). Unlike add-feature, add-item does NOT reject a duplicate name — intentional, so you can author stacks/copies; de-dupe yourself if you need uniqueness.
Target: actorIdentifier embeds on that actor; omit it to create a reusable world Item (optionally in folder). This authors documents — it does NOT roll, equip-in-combat, or spend charges. For features/attacks-as-abilities use add-feature; for free-form system data use create-item / add-feature. To COPY a real item from a compendium (keeps art + stats), prefer import-item.
Inputs: name (required), itemType (required), dex, img, +45 more.
add-journal-imageA
Composition. Append an image page to a journal entry from a Data-relative image path, with an optional caption. GM-only by default; set playerVisible to expose it as a handout.
Inputs: imagePath (required), journalIdentifier (required), caption, pageName, +1 more.
add-region-behaviorA
Add ONE behavior to an EXISTING region — the write create-region only offers at creation time and update-region deliberately never touches. Creates a real RegionBehavior embedded document (validated against the registered behavior types). For teleporters, pass teleportTo {sceneIdentifier, regionIdentifier} and the destination UUID is resolved for you; the landing region's geometry is then sanity-checked and you are WARNED when it contains no grid-snapped token position (the silent teleport no-op: off-grid pads). GM-only.
Inputs: type (required), sceneIdentifier (required), regionIdentifier (required), name, +10 more.
apply-conditionA
[D&D 5e only] Apply or remove one or more conditions on an actor (blinded, frightened, grappled, poisoned, prone, restrained, stunned, unconscious, exhaustion, ...). Set active=false to remove. Exhaustion is leveled — pass exhaustionLevel (1-6; 0 removes). This authors condition state on a creature; it is NOT a combat-automation loop (no duration countdown / save-ends handling). Use list-actors or get-actor to find the actorIdentifier.
Inputs: conditions (required), actorIdentifier (required), active, exhaustionLevel.
asset-infoA
Plane B (file channel, read-only). Report whether a single path under the Foundry Data/ root exists, and (for files) its size, content-type, last-modified, and public HTTPS URL. A cheap existence/metadata check before uploading or linking.
Inputs: remotePath (required).
asset-urlA
Plane B (file channel). Return the public HTTPS URL for a file under the Foundry Data/ root. Pure mapping (no network): everything under Data/ is served at the server root (DESIGN §6), e.g. Data/worlds/w/maps/x.jpg → /worlds/w/maps/x.jpg. Useful for turning an uploaded/known asset path into a link Foundry or a player can load.
Inputs: remotePath (required).
author-npcA
Author a custom NPC (type:npc) from a hand-written stat block — the LAST-RESORT path in the §6 ladder, used ONLY when nothing in the premium MM/PHB/DMG books is a workable base. Prefer create-actor-from-compendium (copy a real Monster Manual creature, optionally with prefab-as-base modifications); if the books are missing what you need, tell the user and ask before authoring rather than inventing content. Prefer the 2024 ruleset (sourceRules:'2024'). Required: name, creatureType (humanoid/undead/beast/dragon/fiend/…), size (tiny…gargantuan), cr (number or fraction string like '1/4'), abilities {str,dex,con,int,wis,cha}, hpAverage, hpFormula (e.g. '5d8+10'), acMode ('default'|'flat'; acValue required if 'flat'). Optional: alignment, savingThrows[], skills[{skill,proficiency}], walk/fly/swim/climb/burrowSpeed, darkvision/blindsight/tremorsense/truesight, damage immunities/resistances/vulnerabilities[], conditionImmunities[], languages[], biography, sourceBook/sourcePage/sourceRules, disposition ('hostile' default | 'friendly' for allies/townsfolk | 'neutral' | 'secret'). Add features, attacks, and spells afterward with add-feature; copy gear from a compendium with import-item.
Inputs: cr (required), name (required), size (required), acMode (required), +31 more.
bulk-deleteA
Permanently delete many world documents of a single type by exact id or exact name. STRICT resolution — no fuzzy/substring matching. IRREVERSIBLE — pass dryRun:true first to preview exactly what would be deleted. For folders use delete-folder. GM-only.
Inputs: identifiers (required), documentType (required), dryRun.
configure-combat-trackerA
Read or configure the combat tracker (the world's core.combatTrackerConfig setting): the custom TURN MARKER shown under the active combatant (enabled / animation / image src / disposition tint), the tracked resource, and skip-defeated. Call with NO arguments to read the current config and the valid animation ids. Changed fields echo previous → new; re-applying the current value is a clean no-op. GM-only.
Inputs: resource, turnMarker, skipDefeated.
configure-dnd5e-settingsA
[D&D 5e] Read or set the dnd5e 6.0 AUTOMATION settings — falling damage, token size ↔ creature size, senses ↔ token vision, exhaustion, initiative grouping, auto-apply Unconscious/Dead at 0 HP, encounter placement, the player damage / effects trays, the bastion system, and the calendar (enabled, daily recovery mode, which calendar). Call with NO arguments to read every switch; pass any of them to change it — each change echoes previous → new, re-applying the current value is a clean no-op, and the switches that need a client reload are named. Allow-listed: nothing else in the settings menu is reachable. GM-only.
Inputs: bloodied, calendar, pietyScore, tokenSizeSync, +17 more.
configure-soundscapeA
Author a scene's atmospheric SOUND SETS — the house module fvtt-mod-soundscape (#6), which does what core Foundry cannot: a POOL of small audio files played at randomized intervals with silence between (a crow, quiet, a distant dog), or overlapped into a seamless crossfaded bed. AmbientSound placeables are positional single-file loops and Playlists have no silence-with-variation, so neither covers this. A scene carries any number of sets, stacked and independent. Actions: "list" (what the scene has, plus what would be playing right now and why a set is idle), "library" (browse the prebaked template catalog by section/category/name), "add" (copy a template by name, or author one from explicit files), "update" (patch one set — named fields only; files replaces the whole pool), "remove" (one set, or "all"). Defaults to the ACTIVE scene when no sceneIdentifier is given. Out-of-range numbers are CLAMPED to the module's limits and the clamp is reported, not applied silently. Audio paths are HEAD-checked: a 404 is kept and warned about (a track has no sensible substitute). Sets are inert data without the module, so this WARNS when it is missing or disabled instead of reporting a working soundscape. GM-only.
Inputs: action (required), name, files, limit, +16 more.
content-auditA
[D&D 5e only] Finishing check for authored content — scan documents for the five strict authoring-quality rules and report violations to fix (read-only; never mutates): • rule 8 — placeholder icons (icons/svg/...) on an actor, item, or authored feature. • rule 7 — GM-fudge / pretend-reskin language in a description or biography ("treat its X as Y", "reflavor", "deals necrotic in place of bludgeoning", "pretend", "is really "). • rule 9 — a magic item on an NPC with no matching world-Item loot twin. • rule 12 — a GM-note / spoiler leaked into a PLAYER-VISIBLE item description ("GM:" asides, "the DM", "fill in the …", "ready-made hook", "to suit your table"). Item descriptions only — an NPC biography is GM-facing, so it is not scanned for this.
RUN THIS before declaring a build done. Target what you built: actorIdentifiers (NPCs, with their gear/features), itemFolders (your loot folder), and/or worldItemIds. With NO target it runs a full sweep of every NPC + every world Item. Fix each finding (set a real icon via update-actor-item/update-item/set-actor-art; replace fudge with real mechanics; mint the missing loot copy; rewrite the item description to innocuous in-world flavor and move the GM note to a GM-only journal) then re-run until clean.
Inputs: itemFolders, worldItemIds, actorIdentifiers.
copy-assetA
Plane B (file channel, write). Copy a file under the Foundry Data/ root over WebDAV; missing destination parent folders are created automatically. (Copying does not affect existing references, so no reference check is needed.) Refuses live world-DB destination paths. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: toPath (required), fromPath (required), overwrite.
create-actor-from-compendiumA
Copy one or more actors from a premium-book compendium pack — the DEFAULT, preferred path for official content (e.g. pull the Owlbear from the Monster Manual). Find the entry with search-compendium / get-compendium-entry, then pass its packId + itemId plus names[] for the new actors. PREFAB-AS-BASE (the §6 step-2 bridge): to make a CUSTOM creature, copy the closest Monster Manual match and pass modifications (update-actor-shaped stat edits — cr/hp/ac/abilities/skills/defenses/biography/currency) to layer onto the world copy in the SAME call; the edits land on the copy only, never the source entry. Pass folder (id or exact name, created if absent) to file the copies directly — no move-documents follow-up. For a fully hand-authored NPC with no compendium base, use author-npc (last resort).
Inputs: names (required), itemId (required), packId (required), folder, +5 more.
create-asset-folderA
Plane B (file channel, write). Create a folder (and any missing parents) under the Foundry Data/ root over WebDAV. Idempotent — succeeds if the folder already exists. Refuses paths inside a live world DB. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: remotePath (required).
create-cardsA
Create a Cards stack (deck, hand, or pile) with optional initial cards. Each card has a name and optional face text (HTML shown on the card — e.g. a Deck of Many Things outcome) and/or img (a Data-relative path), plus a card-level description (GM/meta note). Use for custom themed decks (Deck of Many Things, tarokka, encounter decks). GM-only.
Inputs: name (required), type, cards, folderName, +1 more.
create-drawingsA
Place one or more DRAWINGS (GM annotation shapes: secret-area boxes, trap outlines, zone labels) on a scene. x/y are the TOP-LEFT origin in absolute canvas pixels; pick a shapeType — rectangle/ellipse (width+height), circle (radius), or polygon (flat relative points list). Style with stroke (width/color/alpha), fill (fillType 1 solid / 2 pattern + fillTexture), and an optional centered text label (fontSize/textColor). hidden:true keeps it GM-only; interface:true floats it above fog. The default stroke makes a bare shape visible as an outline. Per-drawing error isolation. Returns created ids. GM-only.
Inputs: drawings (required), sceneIdentifier (required).
create-folderA
Create a sidebar Folder for any world document type, optionally nested under a parent folder of the same type. Use to organize generated content. GM-only.
Inputs: name (required), type (required), color, parentFolder.
create-groupA
Create a dnd5e GROUP actor (type:group) — the shared party stash / travel group. Enrolls world actors as members, can grant every player default ownership (defaultOwnership:'owner' = the classic shared stash), seed shared coin, and optionally crown it the world's PRIMARY PARTY. Member resolution is fail-closed: one bad name and nothing is created. Stock the inventory afterwards with add-item / import-item against the group.
Inputs: name (required), img, members, summary, +5 more.
create-itemA
Create world-level Item document(s) in the Items sidebar — reusable library items (weapons, equipment, consumables, feats, spells). For dnd5e prefer the 2024 data model; pass system-specific data via the "system" field. GM-only. To put items on an actor instead, copy from a compendium with import-item, author one with add-item, or attach raw item data with add-feature (mode "items").
Inputs: items (required), folder.
create-journalA
Create a generic multi-page JournalEntry from caller-supplied pages. Each page is either a TEXT page ({name, content} — HTML, Foundry v13 ProseMirror) or an IMAGE page ({name, kind:"image", src, caption?} — a picture page, e.g. a map legend key), so an image-only journal builds in one call. Unlike create-quest-journal (styled blocks, auto-folders), this takes explicit pages and only folders when folderName is given. Per-page playerVisible exposes a handout; otherwise GM-only.
Inputs: name (required), pages (required), folderName.
create-lightsA
Place one or more AMBIENT LIGHTS (torches, braziers, magical glows) on a scene. x/y are the light CENTER in absolute canvas pixels; dim/bright are radii in grid-distance units (feet), NOT pixels. Set color, alpha (tint intensity), angle (cone), luminosity, attenuation (edge softness), an animation (animationType "torch"/"flame"/"pulse" + speed/intensity for flicker), and a darkness activation range (darknessMin ~0.1 so a torch only lights once the scene dims). walls confines it, vision lets it grant sight. Per-light error isolation. Returns created ids. GM-only.
Inputs: lights (required), sceneIdentifier (required).
create-macroA
Create a world Macro — a hotbar button. type "script" (default) runs JavaScript as the clicking user (e.g. dnd5e.documents.macro.rollItem("Graze") rolls that item on their assigned character); type "chat" posts its text to chat. Optionally grant a player OWNER access and pin the button to their hotbar (hotbarUser + optional hotbarSlot, default first free slot) in the same call — the way to hand a player a one-click ability. GM-only.
Inputs: name (required), command (required), img, type, +3 more.
create-pcA
Build a player character (type:character) headlessly from premium class + species + background by NAME, running real dnd5e advancement so @scale.* (rage damage, sneak attack, breath weapon, …) resolves natively — unlike an NPC. Compendium-first, premium books only, never the SRD (design.md §2.3); a missing class/species/background is an error, not invented. The SKILL owns the math: pass FINAL ability scores (point-buy/array/ASI already applied) and the player CHOICES (skills, fighting style, ancestry…) in choices (level → advancement-id → {chosen|selected|uuid}). Call with no/partial choices first to get a needsChoices[] dry-run (legal options per choice — incl. the available subclasses at level 3 — NOTHING is created); fill the map and re-call. Levels 1-20: HP/features/subclass/spell-slots scale with level (subclass at L3 via a choices uuid; HP per level hpMode avg|max). Multiclass in ONE call via multiclass:[{className,levels}] (className/level is the primary; each multiclass class gets the 2024 proficiency subset; total ≤ 20). Caster spell slots auto-derive from the class; pass spells.cantrips/spells.prepared (names) to add chosen spells. ASI ability-increases ride in the FINAL scores (not applied separately); a feat taken at an ASI tier is added by the skill via add-feature/import-item, like equipment — this tool adds no gear or ASI-feats. If a required advancement (a forced grant / supplied pick / subclass embed) FAILS to apply, the PC is NOT persisted (no junk actor) and success:false is returned with errors[]. Returns {success, actor, applied[], needsChoices[], unresolvedScale[], errors[], warnings[]}.
Inputs: name (required), className (required), level, folder, +9 more.
create-pc-from-prefabA
Create a player character by COPYING a premium-book PREGEN (a complete type:character template — e.g. the PHB class pregens Barbarian…Wizard in dnd-players-handbook.actors, each a ready level-1 build with gear/feats/art) and layering your changes, INSTEAD of building via advancement. The PC family's prefab-as-base path — the §6/§7 analog of create-actor-from-compendium for NPCs, but PC-correct (files under the PC folder, never the NPC one). Resolve the source by prefab NAME (e.g. "Fighter") OR explicit packId+actorId; premium books only, never the SRD (design.md §2.3). Override the pregen's ability array via abilities (final scores) and/or any update-actor-shaped modifications — applied to the COPY only, the source is never touched. @scale resolves natively (it is a real character, no advancement run). Assign the player as owner afterward with set-actor-ownership. Returns {success, from, actor, modificationsApplied, unresolvedScale, warnings}.
Inputs: name (required), folder, packId, prefab, +3 more.
create-playlistA
Create a Foundry Playlist from a list of Data-relative sound paths (e.g. ones just returned by upload-asset). Modes: sequential, shuffle, simultaneous, soundboard. GM-only.
Inputs: name (required), soundPaths (required), fade, mode, +2 more.
create-quest-journalA
Create a multi-page journal (quest log, handout, lore, GM notes) from STRUCTURED typed blocks — a STRUCTURING tool, it never writes the words. You pass pages of blocks (heading / lead / paragraph / readaloud / gmnote / list / grid / html); the tool renders them in the house style and sets per-page visibility (playerVisible -> players can observe a handout; omit -> GM-only). Compose the prose yourself (that's the journal-builder skill's job). For plain raw-HTML pages use create-journal instead.
Inputs: pages (required), title (required), folderName.
create-regionA
Create one or more Regions on an EXISTING scene (the general primitive behind create-teleporter). Each region carries its v14 shapes whole (rectangle/ellipse/polygon in canvas px) plus optional color/visibility/behaviors. Behaviors pass through verbatim: a teleportToken here must already have system.destinations = ["Scene..Region."] (use create-teleporter for the two-new-region convenience, or add-region-behavior to wire one onto an EXISTING region). Returns the created region ids. GM-only.
Inputs: regions (required), sceneIdentifier (required).
create-rolltableA
Create a RollTable from a list of results. Each result is literal text and/or a uuid referencing a REAL premium-book item (rendered as a clickable @UUID link — the way the published loot tables are built; SRD refs are refused). Ranges are auto-assigned from weights (and the formula defaults to 1d) unless you provide explicit ranges/formula. Use for random encounter/loot/rumour/treasure tables. GM-only.
Inputs: name (required), results (required), formula, folderName, +3 more.
create-sceneA
Create a Foundry Scene from a Data-relative background image path (e.g. an uploaded map). Width/height auto-detect from the image when omitted. Places it in a folder (created if absent) and sets navigation (false = a DM-only scene off the player nav bar) in the same call. AUTO-GENERATES the navigation thumbnail from the background (Foundry-native) when no explicit thumb is given — no more thumbnail-less scenes. Optionally set grid size/type/distance/units/color/alpha, token vision, fog mode, lighting (darkness, global light, or a whole environment{}/fog{} mood object + saved camera for pack imports), weather, a linked playlist/journal, a nav thumbnail, padding, provenance flags, and activate it. Can also IMPORT walls + ambient lights from a map sidecar JSON (the walls/lights arrays many battlemaps ship alongside the image): pass them and they are placed on the new scene (legacy or v14 shapes both accepted, normalized to v14). GM-only.
Inputs: name (required), backgroundPath (required), fog, flags, +26 more.
create-scene-notesA
Place map-note PINS on a scene, each linked to a JournalEntry (and optionally a specific page) — the deterministic half of the legend→GM-room-pins feature. Pass absolute canvas pixel x/y (see get-scene-dimensions for the padding-aware math), an optional label/icon/size, and the journal id|name. Per-note error isolation: a pin whose journal does not resolve is reported and skipped, not fatal. GM-only secrecy is the linked journal's ownership, not the pin; global only controls fog occlusion. Returns each created note id (for update-note/delete-note). GM-only.
Inputs: notes (required), sceneIdentifier (required).
create-soundsA
Place one or more positional AMBIENT SOUNDS on a scene (a crackling hearth, a waterfall, dripping cave water) from Data-relative audio paths. x/y are the emitter CENTER in absolute canvas pixels; radius is in grid-DISTANCE units (feet), NOT pixels. Optionally set volume, repeat (loop), walls (muffle through walls), easing (fade by distance), a darkness activation range (night-only sounds), and listener effects (baseEffect/muffledEffect, e.g. "lowpass"). A 404 audio path keeps the path but warns. Distinct from a scene playlist: this is a point emitter players walk into. Returns created ids. GM-only.
Inputs: sounds (required), sceneIdentifier (required).
create-teleporterA
Create a two-way (or one-way) region TELEPORTER between two points on existing scenes — the thing create-scene can only do at import time. Give a CENTER point (canvas px) on each scene (from/to, may be the same scene); a rectangle trigger is placed at each (sized in whole grid cells, grid-snapped by default) and a teleportToken behavior on each points at the OTHER — so a token that walks onto one is sent to the other. Both regions are created before either link is wired (the destination-UUID chicken-and-egg). twoWay:false makes it one-directional. By default the player is ASKED before moving (confirmation dialog — the house pattern for map transitions); confirm:false makes it fire silently (traps/plot teleports only). Regions default to GM/Regions-layer visibility (no player-visible overlay). GM-only.
Inputs: to (required), from (required), color, toName, +6 more.
create-tilesA
Place one or more TILES (props, roof/overhead pieces, decals, video overlays) on a scene from Data-relative image paths. A tile's on-map SIZE is width/height in canvas pixels; x/y are the absolute-canvas-pixel TOP-LEFT (see get-scene-dimensions for padding-aware cell→px math; the tool converts to v14's center-anchored doc coords — never pre-add width/2 yourself). Optionally set rotation, alpha, elevation, sort, texture tint/fit/scale, roof occlusion (occlusionMode: 1 fade / 4 radial so it fades when a token walks under), light/weather restrictions, video loop/autoplay/volume, hidden, locked. Per-tile error isolation; a 404 texture keeps the path but warns. Returns created ids. GM-only.
Inputs: tiles (required), sceneIdentifier (required).
create-wallsA
Create one or more WALL segments on a scene — surgical additions (block a corridor, add a door/secret door) to walls normally drawn in the app or shipped by a map pack. Each wall is a segment x0,y0→x1,y1 (or c:[4]) in absolute canvas pixels. Channels: move (0/20), light/sight/sound (0 none / 10 limited / 20 normal / 30 proximity / 40 distance — omitted channels default to 20 blocking), dir (one-way), door (1 door / 2 secret) + ds (state) + doorSound, and proximity thresholds. Per-wall error isolation. Returns created ids. GM-only.
Inputs: walls (required), sceneIdentifier (required).
delete-actorA
Permanently delete one or more world actors (NPCs/characters) by exact name or ID. IRREVERSIBLE — Foundry has no undo for document deletion; the actor is removed from the world directory. GM-only. Resolution is STRICT (exact id or exact name — no fuzzy matching), so look up the precise name/ID with list-actors first. If the deletion empties a folder the bridge itself created (e.g. "Foundry MCP Creatures"), that folder is auto-removed unless removeEmptyFolder is false.
Inputs: identifiers (required), removeEmptyFolder.
delete-assetA
Plane B (file channel, write). Delete a file under the Foundry Data/ root over WebDAV. REFERENCE-AWARE: consults find-asset-references first and REFUSES if any scene/actor/journal/playlist still points at it (pass force:true to override). Deleting a directory requires recursive:true. Refuses live world-DB paths. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: remotePath (required), force, recursive.
delete-cardsA
Permanently delete one or more Cards stacks by exact id or exact name. STRICT resolution — no fuzzy/substring matching. GM-only.
Inputs: identifiers (required).
delete-chat-messagesA
Delete chat messages: by exact id(s) (a single id is an array of one), or all messages older than a timestamp (beforeTimestamp + confirm:true — handy for the known Molten big-log perf drag), or the entire log (clearAll + confirm:true). Both bulk modes need confirm:true. IRREVERSIBLE. GM-only.
Inputs: ids, confirm, clearAll, beforeTimestamp.
delete-drawingsA
Delete one or more Drawings from a scene by id (from list-drawings). Missing ids are reported, never fatal. GM-only.
Inputs: drawingIds (required), sceneIdentifier (required).
delete-folderA
Permanently delete a folder by exact name or ID. GM-only, IRREVERSIBLE. By default refuses to delete a folder that still contains documents or subfolders (safe for cleaning up empty leftover folders). Pass deleteContents:true to delete the folder AND everything inside it. Defaults to Actor folders; set type for other document folders.
Inputs: identifier (required), type, deleteContents.
delete-itemA
Permanently delete one or more world-level Item documents (Items sidebar) by exact id or exact name. STRICT resolution — no fuzzy/substring matching, so it never deletes the wrong item. GM-only. To remove an item embedded on an actor instead, use remove-from-actor.
Inputs: identifiers (required).
delete-journalA
Permanently delete one or more JournalEntry documents by exact id or exact name. STRICT resolution — no fuzzy/substring matching. GM-only.
Inputs: identifiers (required).
delete-journal-pageA
Delete ONE page from a JournalEntry by page id (from list-journals), leaving the rest of the entry intact. Use to remove a stray/mistaken page instead of deleting and rebuilding the whole journal. GM-only.
Inputs: pageId (required), journalId (required).
delete-lightsA
Delete one or more AmbientLights from a scene by id (from list-lights). Missing ids are reported, never fatal. GM-only.
Inputs: lightIds (required), sceneIdentifier (required).
delete-macroA
Delete world macros by id or exact name, scrubbing any user hotbar slots that pointed at them so no dead buttons are left behind. GM-only.
Inputs: macros (required).
delete-noteA
Remove one or more map-note pins from a scene by note id (from create-scene-notes). Missing ids are reported, never fatal. GM-only.
Inputs: noteIds (required), sceneIdentifier (required).
delete-playlistA
Permanently delete one or more Playlist documents by exact id or exact name. STRICT resolution — no fuzzy/substring matching. GM-only.
Inputs: identifiers (required).
delete-regionA
Delete one or more Regions from a scene by id. Missing ids are reported, never fatal — and if a surviving teleporter elsewhere still points at a deleted region, that orphan is warned. GM-only.
Inputs: regionIds (required), sceneIdentifier (required).
delete-rolltableA
Permanently delete one or more RollTable documents by exact id or exact name. STRICT resolution — no fuzzy/substring matching. GM-only.
Inputs: identifiers (required).
delete-sceneA
Permanently delete one or more Scene documents by exact id or exact name. STRICT resolution — no fuzzy/substring matching. GM-only.
Inputs: identifiers (required).
delete-soundsA
Delete one or more AmbientSounds from a scene by id (from list-sounds). Missing ids are reported, never fatal. GM-only.
Inputs: soundIds (required), sceneIdentifier (required).
delete-tilesA
Delete one or more Tiles from a scene by id (from list-tiles). Missing ids are reported, never fatal. GM-only.
Inputs: tileIds (required), sceneIdentifier (required).
delete-tokensA
Remove one or more PLACED tokens from a scene by token id (from list-tokens) — clears the map instance only; the sidebar actor survives (delete-actor removes that). Missing ids are reported, never fatal. GM-only.
Inputs: tokenIds (required), sceneIdentifier (required).
delete-wallsA
Delete one or more Walls from a scene by id (from list-walls) — e.g. open up a sealed passage. Missing ids are reported, never fatal. GM-only.
Inputs: wallIds (required), sceneIdentifier (required).
disconnect-bridgeA
Log the DM Assistant bridge user out of the live Foundry world: closes the persistent headless browser session so the user goes inactive (drops off the active-player list). The world itself keeps running and the MCP server stays up — the NEXT tool call transparently reconnects (wake → join → ready), so nothing needs to be restarted. Use at the end of a work session ("log out please"); safe to call any time — already disconnected is a clean no-op.
Inputs: no named parameters.
download-assetA
Plane B (file channel, read-only). Download a file from under the Foundry Data/ root (over WebDAV) to a local path on this machine. For grabbing an existing asset to inspect or re-process.
Inputs: localPath (required), remotePath (required).
duplicate-actorA
Clone one or more existing WORLD actors as full copies — the whole sheet travels (system data, embedded items, spells, effects, prototype token), so every copy stays fully rollable. THE sandbox path: clone PCs as "(Sim)" copies (suffix: " (Sim)") so a player can re-run a battle without touching the real sheets. Resolution is STRICT (exact id or exact name — look up with list-actors first); a missing source is reported per-actor, never fatal to the batch. Name the copies with newNames[] (index-aligned) or a suffix on every source name (default " (Copy)"). Pass folder (id or exact name, created if absent) to file the copies together — omit to file each beside its source. Pass owner (user id or exact/partial name; + ownershipLevel, default OWNER) to set each copy's ownership to exactly {default: NONE, that user: level} — omit to copy the source's ownership unchanged. GM-only.
Inputs: actorIdentifiers (required), owner, folder, suffix, +2 more.
export-actorA
Write ONE actor to a LOCAL file as a full-fidelity native Foundry JSON export — the complete document source (system, every embedded item with its uses/charges and flags, effects, prototype token, ownership), stamped with the exportSource envelope so the file round-trips through the sheet's Import Data button. THE character backup/restore path (get-actor is a lossy summary view). Accepts a placed token id to capture an unlinked token instance. Refuses to overwrite an existing file unless overwrite:true.
Inputs: localPath (required), identifier (required), overwrite.
export-chat-logA
Export the chat transcript to a LOCAL absolute file AND/OR a WebDAV Data/ path (returns its public URL). Formats: markdown | html | json | plaintext. Refuses to overwrite an existing file at either destination unless overwrite:true. WebDAV needs MOLTEN_WEBDAV_PASSWORD.
Inputs: limit, format, localPath, overwrite, +2 more.
find-asset-referencesA
Reference integrity. Find every world document (scenes, actors, items, journals, playlists, macros, roll tables) that references a given asset path under Data/. Use this BEFORE deleting or moving a file to see what would break. Read-only.
Inputs: paths (required).
get-actorA
Retrieve D&D 5e character information optimized for minimal token usage. Returns: full stats (abilities, skills, saves, AC, HP, 2024 weapon-mastery kinds), action names, active effects/conditions (name only), and ALL items with minimal metadata (name, type, equipped status, attunement, weapon mastery property) without descriptions. Perfect for checking equipment or identifying what to investigate further. Use get-actor-entity to fetch full details for specific items, spells, or effects.
Inputs: identifier (required).
get-actor-entityA
Retrieve full details for a specific entity from a character. Works for items (feats, equipment, spells), actions (strikes, special abilities), or effects/conditions. Returns complete description, all system data, and (for items) module flags — the read path for flag forensics like item-piles transfer residue. Use this after get-actor when you need detailed information about a specific entity.
Inputs: entityIdentifier (required), characterIdentifier (required).
get-combat-statsA
Fold Battle Flow's stat-stamped chat messages into per-combat analytics: damage dealt/taken (with per-round rate), healing + overheal, verdict-flip credits, spend economy (resource pools AND spell slots), buff-die (Bless) margin flips, and session flavor (nat 20s/1s, advantage economy, death saves). Read-only scan; reverted applications are subtracted; unlinked monsters aggregate by archetype. GM-facing — filter at call time via sections/actor/combat/since. includeLedger:true appends the folded JSON for charts.
Inputs: actor, since, combat, sections, +1 more.
get-compendium-entryA
Retrieve a specific compendium entry (monster, item, spell, etc.) by pack id + entry id. Returns the full stat block — items, spells, abilities, effects, system data — needed for actor/item creation. Set compact=true for a condensed stat block when full detail is not needed. An SRD (dnd5e.*) pack id is refused — author only from the premium books (design.md §2.3).
Inputs: itemId (required), packId (required), compact.
get-current-sceneA
Get information about the currently active scene, including tokens and layout
Inputs: includeHidden, includeTokens.
get-groupA
Read a dnd5e group actor the group-shaped way: member roster (with dangling ids flagged), shared currency, shared inventory, ownership (default + per-user), and whether it is the primary party. Use this instead of get-actor for type:group.
Inputs: groupIdentifier (required).
get-itemA
Retrieve a single world-level Item document with its full system data, embedded effects, flags, and flattened description. Resolves by id (most reliable), exact name, or case-insensitive name. Use list-items first to find ids. For an UNIDENTIFIED dnd5e item (system.identified=false) name is the mystery mask; the real source name is returned as trueName.
Inputs: identifier (required).
get-rolltableA
Read a RollTable's FULL contents — every entry with its roll range, weight, drawn flag, the result text (HTML/@UUID enrichers intact), and any linked items surfaced as uuid + label — sorted low-to-high so a d table reads 1..N. The deterministic way to inspect or audit a table's entries without brute-force rolling (list-rolltables gives only a per-table summary; roll-on-table draws one random entry). Resolves by id or exact name.
Inputs: identifier (required).
get-scene-dimensionsA
Read a scene's live PADDED-CANVAS geometry (by id or exact name): total width/height, the background rect within the padding (sceneX/sceneY/sceneWidth/sceneHeight), grid size/distance, and rows/columns. A scene insets its background by a padding border, so a placeable's canvas pixel is NOT just gridCell×size — use sceneX/sceneY to offset. Feeds the legend→pins cell→px math. Works on any scene (no need to activate it).
Inputs: sceneIdentifier (required).
get-world-infoA
Get basic information about the Foundry world and system, including the world description (HTML) and join-page background path. READ-ONLY: world metadata has no write tool (the /setup editWorld route needs a role-4 GAMEMASTER session; the bridge user is deliberately an ASSISTANT) — hand edits go through the in-app "Edit World" dialog.
Inputs: no named parameters.
import-cardsA
Instantiate a core Foundry PRESET deck into the world (e.g. "pokerDark"/"pokerLight" — a standard 52-card deck). Cards have no premium-book compendium, so this is the ready-made deck path; build themed D&D decks with create-cards. GM-only.
Inputs: preset (required), name, folderName.
import-itemA
[D&D 5e only] COPY an existing item from a compendium pack onto an actor (or into the world Items sidebar), keeping its artwork, full system data, and activities. PREFER THIS over add-item for any real piece of gear — a plain greatsword, a Potion of Healing, a +1 shield, a magic weapon: copying brings the correct PHB/DMG 2024 stats AND the graphic, where authoring from scratch does not.
WORKFLOW: 1) find the item with search-compendium (prefer the 2024 packs: "dnd-players-handbook.equipment", "dnd-dungeon-masters-guide.equipment" — premium books ONLY, never the dnd5e.* SRD); 2) import-item with its packId + itemId; 3) for a CUSTOM item, copy the closest base then refine it with update-actor-item / manage-activity / manage-effect and rename via name.
Optional on-copy tweaks: name (rename), quantity, equipped, identified, container (nest in a bag/chest), folder (world target only). Target an actor with actorIdentifier, or omit it to build a reusable world Item. Use add-item only for genuine homebrew with no compendium base.
Inputs: itemId (required), packId (required), name, folder, +7 more.
import-rolltableA
Copy a whole RollTable from a compendium pack into the world (e.g. a DMG treasure / magic-item table). Roll tables are world-only at roll time, so a published table must be imported before roll-on-table can use it; the embedded results — including their @UUID item links — come along intact. Premium-book packs only (SRD refused). GM-only.
Inputs: itemId (required), packId (required), folderName.
inspect-pc-advancementA
Read-only: report the player CHOICE points a premium class exposes up to a level — each advancement's id, type (Trait/ItemChoice/Subclass), how many to pick, and the legal options — so the skill can ask the DM and fill create-pc's choices map without inventing anything. Resolve by className OR classUuid (exactly one); premium books only, never the SRD. Touches no actor.
Inputs: level, className, classUuid.
level-up-pcA
Add ONE level to an existing PC (type:character) and apply that level's advancement IN PLACE. Same className as a class the PC already has → a single-class level-up; a class it does NOT have → a MULTICLASS add (the PC gets the 2024 multiclass proficiency SUBSET, not the full first-level kit). HP/features/subclass(@ the class's level 3)/spell-slots scale; @scale stays native. Like create-pc: call with no/partial choices to get a needsChoices[] dry-run (e.g. the subclass options at level 3 — the actor is NOT touched); fill choices (level → advancement-id → {chosen|selected|uuid}) and re-call. ASI ability bumps are NOT applied here — raise the final scores with update-actor; a feat taken at an ASI tier is added with add-feature. If a required advancement FAILS to apply, the PC is rolled back to its prior level and success:false is returned with errors[]. Required: actorIdentifier, className. Returns {success, actor (incl. classLevel + classes[]), applied[], needsChoices[], unresolvedScale[], errors[], warnings[]}.
Inputs: className (required), actorIdentifier (required), hpMode, choices, +1 more.
link-quest-to-npcA
Append a link from a quest journal to a REAL world NPC: resolves the actor, inserts a Foundry @UUID[Actor.id]{Name} enricher link (clickable on render) in a GM note, labelled with the relationship. Refuses an unknown NPC (no dead links) — create the actor first.
Inputs: npcName (required), journalId (required), relationship (required), pageId.
list-actor-ownershipA
List current ownership permissions for actors, showing which players have what access levels. Each row reports the EFFECTIVE level plus its source: 'explicit' (the actor's ownership map holds an entry for that player) or 'inherited' (the actor's default, also reported per actor as defaultPermission). Use it to tell an explicit NONE — a stored deny that overrides a permissive default — from a player who simply has no entry.
Inputs: actorIdentifier, playerIdentifier.
list-actorsC
List all available characters with basic information
Inputs: type.
list-assetsA
Plane B (file channel, read-only). List the immediate contents of a directory under the Foundry Data/ root over WebDAV (folders + files, with size / type / public URL). Use to browse uploaded assets, e.g. worlds/your-world/assets/audio. Empty/omitted path lists the Data/ root.
Inputs: remotePath.
list-cardsB
List Cards stacks with id, name, type (deck/hand/pile), and card count.
Inputs: no named parameters.
list-chat-messagesA
List recent chat messages (id, author, time, whisper/blind, content preview). Use to find ids for delete, verify a post, or preview before export. contentMode:"none" keeps it cheap on a huge log.
Inputs: limit, contentMode, sinceTimestamp.
list-compendium-packsA
List the available compendium packs. SRD (dnd5e.*) packs are excluded — only the premium book packs (and any other non-SRD packs) are listed (design.md §2.3).
Inputs: type.
list-drawingsA
List every Drawing on a scene — id, origin (x/y), shape type + dimensions (width/height/radius/pointCount), rotation, label text, fill type, stroke color, hidden/locked/interface. Read-only; the inspect step before update-drawings / delete-drawings.
Inputs: sceneIdentifier (required).
list-foldersA
Read the sidebar folder TREE — every world folder (or one document type) in tree order with its id, nesting depth + "/"-joined path, hex color, raw sort value, parent, direct document count, and subfolder count. Siblings are listed ALPHABETICALLY — which matches how the v14 sidebar actually renders FOLDERS (it ignores sort for them), so this listing order IS the sidebar order. The inspect step the folder tools were missing: find the ids/names for update-folder (rename/recolor/reparent), delete-folder, move-documents, and the folder params on the create tools — without guessing what exists. Read-only.
Inputs: type.
list-itemsA
List world-level Item documents, optionally filtered by type, name substring, or folder.
Inputs: type, folder, nameFilter.
list-journalsA
List all journal entries, or read a specific journal/page. Without parameters: lists all journals with their pages (id, name, type). With journalId: reads the journal's first text page content and shows all available pages. With journalId + pageId: reads a specific page's full content.
Inputs: pageId, journalId, filterQuests, includeContent.
list-lightsA
List every AmbientLight on a scene — id, center (x/y), rotation, dim/bright radii, color, cone angle, animation type, hidden, walls/vision. Read-only; the inspect step before update-lights / delete-lights.
Inputs: sceneIdentifier (required).
list-macrosA
List every macro in the world: id, name, type (script/chat), author, icon, a command preview, and every user hotbar slot it is pinned to. The read to run before delete-macro.
Inputs: no named parameters.
list-notesA
List every MAP-NOTE PIN on a scene — id, position (x/y), label text, linked journal entryId/pageId, icon src + size, fog global, font. Read-only; the inspect step that feeds update-note / delete-note (create-scene-notes places them).
Inputs: sceneIdentifier (required).
list-playlistsA
List Playlist documents with id, name, mode, track count, and whether each is currently playing.
Inputs: no named parameters.
list-regionsA
List every Region on a scene — id, name, each shape's bounds, and any teleporter destinations. Read-only; use it to find region ids for update-region / delete-region.
Inputs: sceneIdentifier (required).
list-rolltablesB
List RollTable documents with id, name, formula, result count, and description.
Inputs: no named parameters.
list-scenesA
List Scene documents with id, name, active flag, dimensions, grid size, and background path. Optionally filter by name substring or show only the active scene.
Inputs: filter, includeActiveOnly.
list-soundsA
List every AmbientSound on a scene — id, name, center (x/y), radius, audio path, volume, repeat/walls/easing flags, darkness range, base effect. Read-only; the inspect step before update-sounds / delete-sounds.
Inputs: sceneIdentifier (required).
list-tilesA
List every Tile on a scene — id, position (x/y = the TOP-LEFT in canvas pixels, converted from v14's center-anchored doc coords), size (width/height), rotation, elevation, sort, texture src, image scale, hidden/locked. Read-only; the inspect step before update-tiles / delete-tiles (you need the ids + current values to edit).
Inputs: sceneIdentifier (required).
list-tokensA
List every PLACED TOKEN on a scene (by id or exact name — any scene, not just the active one) — id, name, position (x/y), size, rotation, elevation, hidden, disposition, actorId, art src + scale, lockRotation. Read-only; the inspect step that feeds update-token / delete-tokens. The token ids also work as the actorIdentifier of the actor tools (get-actor, update-actor, update-actor-item, remove-from-actor, add-item/add-feature, import-item, manage-activity/-effect, apply-condition) — targeting a token id edits THAT instance's own delta, the way to re-gear/wound ONE placed copy of an unlinked NPC (base-actor edits never reach tokens already on a scene).
Inputs: sceneIdentifier (required).
list-usersA
List every user account in the world: id, name, role (player/trusted/assistant/gamemaster), whether they are currently connected, player color, pronouns, avatar, their assigned character, and their LANDING SCENE if one is set (where they come up at login — shown only when assigned, since an unassigned user just follows the active scene). The read to run before update-user and set-landing-scene.
Inputs: no named parameters.
list-wallsA
List walls on a scene — id, segment c:[x0,y0,x1,y1], move/sight/light/sound channels, one-way dir, door kind + state + sound. A populated scene carries HUNDREDS of walls: pass doorsOnly:true to get just the doors (the usual edit loop). Read-only; the inspect step before update-walls / delete-walls.
Inputs: sceneIdentifier (required), doorsOnly.
manage-activityA
[D&D 5e only] Add / edit / remove / list Activities on an item — the rollable things (attack, damage, save, heal, check, utility, cast). Target an item on an actor (set actorIdentifier) or a world item (omit it). This authors actions like a Multiattack (action="add", type="utility", name="Multiattack"), a heal, an ability-check, a saving-throw activity, OR a spell-casting item (action="add", type="cast", spellUuid=…, charges=…, saveDC/attackBonus=… to pin a fixed challenge) — the cast LINKS a real compendium spell so its measured template + save/attack fire for free. appliesEffects puts real ActiveEffects on the targets (dnd5e 6.0) — e.g. [{ref:"Poisoned", onSave:false}] on a save activity; they inherit the activity duration/expiry. Use action="list" (or get-actor-entity) to find activityIds, then edit/remove by id. EDIT changes only name, duration, template, affects, behaviors and appliesEffects — every other field goes through patch (dot-paths relative to the activity, e.g. {"attack.bonus":"3"}) or a remove + add. Authoring only — it does not run combat.
Inputs: action (required), itemIdentifier (required), name, type, +32 more.
manage-calendarA
[D&D 5e] The in-world date and time (dnd5e 6.0 calendar on Foundry v14 game.time). action "read" (default) reports the calendar in use, whether it is enabled, and the current date / time (year as displayed, month name + number, day, weekday, hour, minute, plus the system's own formatted / approximate strings). "advance" moves time by rounds / minutes / hours / days / seconds (negative rewinds) — this is what triggers dawn / dusk / day recovery and bastion turns when the calendar is enabled. "set" jumps to a date (year / month by number or name / day) and/or a time of day (hour / minute); omitted parts keep their value. Writes report before → after. Enable the calendar / pick one with configure-dnd5e-settings. GM-only writes.
Inputs: day, days, hour, year, +7 more.
manage-effectA
[D&D 5e] Create / edit / delete / list ActiveEffects on an actor or an item. Effects carry changes ({key, value, type}) that modify the target — e.g. +1 AC ({key:"system.attributes.ac.bonus", value:"1", type:"add"}) or resist fire — or, with a dnd5e 6.0 RULES type, modify a ROLL: type "dnd5e.advantage" | "dnd5e.bonus" | "dnd5e.minimum" | "dnd5e.maximum" on key "attack" | "check" | "d20" | "save" (damage / healing: bonus only), e.g. +1d4 on History checks = {key:"check", value:"1d4", type:"dnd5e.bonus", conditions:{k:"roll.skill", v:"his"}}, disadvantage on STR d20 tests = {key:"d20", value:"-1", type:"dnd5e.advantage", conditions:{k:"roll.ability", v:"str"}}. 6.0 conditions (a Filter) gate the whole effect ("while Bloodied") or one change; duration.expiry takes the combat / rest / source-target turn events. Target the actor (actorIdentifier), an embedded item (actorIdentifier + itemIdentifier), or a world item (itemIdentifier alone). Use action="list" to find effectIds (it reads conditions + rules back). Item effects transfer to the owning actor by default. Authoring only — it sets effect data, it does not run combat.
Inputs: action (required), name, patch, changes, +10 more.
manage-group-membersA
Add and/or remove members on an existing dnd5e group actor. Uses the system addMember/removeMember API; every requested actor gets a reported outcome (added / removed / skipped with reason) and the final roster is echoed back.
Inputs: groupIdentifier (required), add, remove.
move-assetA
Plane B (file channel, write). Move/rename a file under the Foundry Data/ root over WebDAV; missing destination parent folders are created automatically. REFERENCE-AWARE: by default REFUSES with a report if anything references the source (moving would break those pointers). Pass relink:true to move AND rewrite all references (old→new), or force:true to move without relinking. Refuses live world-DB paths. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: toPath (required), fromPath (required), force, relink, +1 more.
move-documentsA
Move one or more world documents of a single type into a target folder (resolved by id or name; created at root if absent). Pass an empty targetFolder to move them to the root. GM-only.
Inputs: identifiers (required), documentType (required), targetFolder.
place-tokensA
Place one or more actors' tokens on a scene (batch encounter prep — e.g. drop the whole hobgoblin band on the bridge). Each entry names an actor (id or EXACT name) + an absolute canvas-pixel x/y; the token is built from the actor's PROTOTYPE (so the house token defaults — auto-rotate, ring, disposition — carry over), with optional per-copy hidden/elevation/rotation/name/disposition overrides. Repeat an actor for several copies. The GM can always drag tokens in the app instead — this is for scripted/batch placement. GM-only.
Inputs: tokens (required), sceneIdentifier (required).
post-item-cardA
Post a rich dnd5e card for an actor's item/feature/spell with WORKING buttons (Attack/Damage/Apply-Effects), or roll an attack/damage to chat. Drives the dnd5e Activity system — the only way to get interactive buttons without a module. Items with no activity return a clear reason (use send-chat-message for a plain card). GM-only.
Inputs: item (required), actor (required), action, consume, +2 more.
pull-users-to-sceneA
Pull specific users' VIEW to a scene WITHOUT changing which scene is active — the party-split path (one player off in a side scene while everyone else stays put). This is what a cross-scene teleporter does as a side effect of moving a token; here it is the whole operation, so no token has to move. Viewing needs NO scene ownership — Scene#view() has no permission gate, which is why players already move between scenes they do not own. ⚠️ Only CONNECTED users can be pulled: core silently skips offline ones, so this reports pulled vs offline vs notFound per user instead of a blanket success. ⚠️ It also cannot pull the BRIDGE user itself — pullUsers rides a socket emit, which is never echoed back to the sending client (reported as selfSkipped). Look up names with list-users. GM-only.
Inputs: sceneIdentifier (required), userIdentifiers (required).
read-packA
Read a Tom-Cartos-style Foundry SCENE-PACK MODULE off disk (a module.json + LevelDB/NeDB compendium packs — Scene/JournalEntry packs AND modern type:"Adventure" packs whose single Adventure doc embeds every scene/journal) and return its era-normalized documents for import. OFF-LINE and Node-only: it reads files, never the live world. Detects the pack's Foundry era from field shape (older v10/NeDB vs newer v13/LevelDB), extracts each Scene (dimensions, grid, background, thumbnail, walls, lights, regions/teleporters) and JournalEntry (pages), strips cli pack artifacts, and — when given the destination root the skill chose — emits per-asset path REWRITE HINTS (the module-relative %-encoded src → a clean Data-relative path). Also discovers any standalone TILE images the pack ships (building/prop pieces with a Tile_<W>x<H> grid footprint in the name, not referenced by any scene) so the skill can make them available for the GM to drop onto scenes. The heavy per-scene walls/lights/regions are written to PAYLOAD FILES and referenced by placeablesPath (NOT inline — the response cap truncates them at scene scale); pass that path to create-scene, which reads it server-side. The skill then uploads the assets, recreates the scenes/journals, and (modern packs) remaps the cross-scene teleporters. Handles all eras: modern v13/LevelDB (via the @foundryvtt/foundryvtt-cli child process) AND legacy v10/NeDB .db (parsed directly — no cli needed). The full manifest is PAGED to fit the response cap: returns totalScenes + a page of scenes + nextOffset (call again with offset until null). Pass index:true first for a tiny names-only survey of the whole pack (variant planning + dedup) before paging the heavy import.
Inputs: modulePath (required), index, offset, destRoot, +2 more.
relink-assetA
Reference integrity. Rewrite every reference from one asset path to another (e.g. after moving/renaming a file) so nothing breaks. Pass dryRun:true to preview the documents that would change without writing. GM-only.
Inputs: newPath (required), oldPath (required), dryRun.
remap-teleportersA
Second pass of a scene-pack import: rewrite cross-scene teleporter destinations after the scenes + regions have been created. A pack teleporter points at Scene..Region., but the import mints FRESH ids, so every destination is stale until remapped. Pass the import sourceModule; this reconstructs the old→new scene/region id maps from the provenance flags the scenes + regions carry and rewrites every teleportToken destination. Idempotent (safe to re-run), and reports destinations that point outside the import (e.g. a variant you skipped) rather than dropping them silently. Call it ONCE after all chosen scenes are imported. GM-only.
Inputs: sourceModule (required).
remove-from-actorA
Delete items already on an actor, identified by itemIds and/or itemNames (optionally constrained by type). GM-only. Use get-actor to find item ids.
Inputs: actorIdentifier (required), type, itemIds, itemNames.
request-rollA
Post a click-to-roll request card (saving throw / ability check / skill) that players click to roll their OWN check — the table-facing "everyone make a DEX save (DC 15)" prompt. Uses the dnd5e inline roll enricher. GM-only.
Inputs: kind (required), dc, skill, flavor, +2 more.
roll-on-tableA
Roll on a world RollTable and return the drawn result(s). Evaluates without marking results drawn or posting to chat. Any @UUID item links in a drawn result are surfaced as importable (uuid + label) so loot can be pulled into the world. (World tables only — copy a compendium/DMG table in first with import-rolltable.)
Inputs: identifier (required).
screenshot-sceneA
Render a scene in the headless bridge and capture a PNG to a local file — visual QA for imports/maps. Views the scene, waits for the WebGL canvas to draw, fits the whole map into the viewport (or keeps the saved camera with fit:false), and optionally draws numbered markers over each map-note pin (mark:true) to check legend-pin placement (a view-only overlay, no document changes). Returns the file path + scene metadata; open/read that file to view the image. GM-only.
Inputs: sceneIdentifier (required), fit, mark, outputPath.
search-actor-contentsA
Search within a character's items, spells, actions, and effects. More token-efficient than get-actor when you need specific items. Supports text search (name/description) and type filtering. Returns matching items with full details including targeting info for spells. Use this to find specific spells, equipment, feats, or abilities without loading the entire character.
Inputs: characterIdentifier (required), type, limit, query, +1 more.
search-compendiumA
Broad NAME search across the premium book compendium packs (any document type). The SRD (dnd5e.*) packs are NOT searched and never appear in results — the authoring library is the premium books only (design.md §2.3). Matches entity NAMES only (all whitespace-separated terms must appear); descriptions and traits are NOT searchable. Premium-first ranked, exact-name first. For faceted discovery by real system data (CR/type/size, spell level/school, item rarity/type), use the type-specific tools instead: search-compendium-creatures, search-compendium-spells, search-compendium-items. Use this for a quick name lookup, then inspect with get-compendium-entry.
Inputs: query (required), limit, packType.
search-compendium-creaturesA
D&D 5e CREATURE DISCOVERY: find creatures matching faceted criteria (Challenge Rating, type, size, spellcasting, legendary actions) across the premium book Actor packs only — the SRD (dnd5e.*) packs are excluded and never appear in results (design.md §2.3). Backed by the system Compendium Browser, so CR/type/size check real system data (not name heuristics); hasSpells/hasLegendaryActions are approximate index flags. Returns minimal hits ({id,name,type,uuid,pack,packLabel,img,facets}) premium-first ranked — identify candidates by name, then pull full stat blocks with get-compendium-entry. High result limits for complete encounter-building surveys.
Inputs: size, limit, hasSpells, creatureType, +2 more.
search-compendium-itemsA
D&D 5e ITEM/GEAR DISCOVERY: find equipment, weapons, armor, consumables, and treasure matching faceted criteria (rarity, subtype, properties, magical, name) across the premium book packs only — the SRD (dnd5e.*) packs are excluded and never appear in results (design.md §2.3). Backed by the system Compendium Browser, so filters check real item data (not name heuristics). Returns minimal hits ({id,name,type,uuid,pack,packLabel,img,facets}) premium-first ranked — identify candidates here, then pull full detail with get-compendium-entry. Use documentType to narrow the item family (gear=all, or weapon/armor/consumable).
Inputs: name, limit, rarity, magical, +3 more.
search-compendium-spellsA
D&D 5e SPELL DISCOVERY: find spells matching faceted criteria (level, school, damage type, name) across the premium book packs only — the SRD (dnd5e.*) packs are excluded and never appear in results (design.md §2.3). Backed by the system Compendium Browser, so filters check real spell data (not name heuristics). Returns minimal hits ({id,name,type,uuid,pack,packLabel,img,facets}) premium-first ranked — identify candidates here, then pull full detail with get-compendium-entry. damageType is a two-stage refine (loads candidate spells to inspect their activities).
Inputs: name, limit, damageType, spellLevel, +1 more.
search-journalsA
Search through all pages of all journal entries for specific content or keywords. Returns which specific page matched, so you can read it with list-journals using journalId + pageId.
Inputs: searchQuery (required), searchType.
send-chat-messageA
Post a message to the Foundry chat log as the GM bridge user. Content is HTML. Choose a visibility mode (public / gm whisper / blind / self) and optionally speak AS a character (speakerActor). Embed images via the images param (local files upload over WebDAV; Data-relative paths and https URLs link directly — uploaded files are PUBLIC). GM-only.
Inputs: content (required), style, enrich, flavor, +5 more.
set-actor-artA
Composition. Set an actor's portrait image, and by default its prototype token art too, from a Data-relative path. The portrait (actor.img) must be a STILL image; pass tokenImagePath to give the prototype TOKEN an animated video (.webm/.mp4) while keeping a still portrait (the JB2A-effect pattern). GM-only.
Inputs: imagePath (required), actorIdentifier (required), applyToToken, tokenImagePath.
set-actor-ownershipA
Set ownership permissions for actors. Use permissionLevel OWNER/OBSERVER/LIMITED to grant access, NONE to explicitly deny it (a stored level-0 entry that OVERRIDES the actor's default ownership), or INHERIT to remove the player's entry so the actor's default applies again. On an actor whose default is permissive (e.g. a shared party stash defaulting to OWNER), NONE revokes access the player would otherwise inherit and INHERIT is what restores it. Supports individual assignments like "Make John the owner of Aragorn" and bulk operations like "Give the party observer access to all friendly NPCs".
Inputs: actorIdentifier (required), permissionLevel (required), playerIdentifier (required), confirmBulkOperation.
set-journal-page-visibilityA
Flip one journal PAGE between player-visible (a handout players can OBSERVE) and GM-only, WITHOUT rewriting its content. Sets the page ownership default. Use this to reveal/hide an existing page — e.g. a page that came up GM-only from an append — instead of rebuilding the whole journal. GM-only.
Inputs: pageId (required), journalId (required), playerVisible (required).
set-landing-sceneA
Assign where a user LOGS IN — a per-user landing scene that persists, so a party split survives a refresh. Core Foundry has no such thing: every user lands on the one ACTIVE scene, and the User document has no scene field to change that (verified on v14.364). This writes a durable User FLAG that the house module fvtt-mod-openserver reads on ready and acts on with Scene#view() — so the module must be installed and enabled for it to do anything, and this tool WARNS when it is not, instead of reporting a working assignment. Unlike pull-users-to-scene, this works on OFFLINE users — that is the whole point. Sticky until cleared: pass sceneIdentifier "none" to put users back on the active scene. Reports who is assigned and who still follows the active scene. Needs NO scene ownership. GM-only.
Inputs: sceneIdentifier (required), userIdentifiers (required).
set-primary-partyA
Read or change the world's dnd5e PRIMARY PARTY (the group shown in the players sidebar; XP awards and party overviews target it). No arguments → report the current primary party. groupIdentifier → point it at that group. clear:true → unset.
Inputs: clear, groupIdentifier.
set-user-avatarA
Set a Foundry user's avatar — the portrait shown next to that user's chat messages. Defaults to the bridge user (MCP-Claude), so this is how you give the MCP's own chat posts a portrait instead of the default mystery-man. Pass a Data-relative path or https URL (upload local files with upload-asset first). GM-only.
Inputs: avatar (required), user.
update-actorA
[D&D 5e only] Edit an EXISTING actor's own stat-block fields. Supply only the groups you want to change: • identity + prototype token — name, tokenName (prototype nameplate ≠ actor name), img, disposition (friend/foe), tokenAutoRotate (face movement / lockRotation), tokenRing (dynamic ring), tokenScale (art size), tokenRotation (facing), tokenDisplayName / tokenDisplayBars (nameplate / HP-bar visibility) — the PROTOTYPE-token editor. (elevation / hidden / x / y are placement-only: edit those on a dropped token with update-token) • details — size, cr*, creatureType*, creatureSubtype*, swarmSize*, alignment, biography, source • abilities — abilities.{str..cha}, savingThrows (replace), skills (merge; proficiency none/proficient/expert) • weaponMasteries [PC] — {mode: replace|add|remove, values: ["greatsword", ...]} — 2024 Weapon Mastery weapon KINDS (base weapon ids, not mastery names) • vitals — hp, ac, initiative • movement, senses • defenses — damageImmunities / damageResistances / damageVulnerabilities / conditionImmunities / languages (each {mode: replace|add|remove, values, custom?}), telepathy • resources* — legendaryActions, legendaryResistances, lair • spellcasting* — {level, ability}: caster level (the sheet's "Spellcasting Level", which DERIVES the slot pools) + casting ability (which drives save DC / attack bonus) • 2024* — habitat, treasure • currency — coins {mode: set|add, pp, gp, ep, sp, cp} (carried money)
Fields marked * are NPC-only (skipped with a warning on player characters). This authors the stat block; it does NOT edit embedded items (use update-actor-item / add-feature / manage-activity) or run combat. Use list-actors or get-actor to find the actorIdentifier.
Inputs: actorIdentifier (required), ac, cr, hp, +37 more.
update-actor-itemA
[D&D 5e] Edit an item embedded on an actor (weapon / feature / spell / equipment). Apply a dot-path patch (values applied as-is; arrays/Sets replace whole) and/or deletePaths (remove keys, e.g. an activity by id), and/or change name/img. This is the low-level item editor — to add/edit/remove activities (attacks, saves, heals, etc.) prefer manage-activity, which knows the shapes. Use get-actor or get-actor-entity to find the item and the exact paths/ids to change.
Inputs: itemIdentifier (required), actorIdentifier (required), img, name, +3 more.
update-drawingsA
Edit one or more placed DRAWINGS by id (from list-drawings): MOVE via x/y, RESIZE via width/height/radius or replace polygon points, restyle stroke/fill, change or clear the text label (text:""), toggle hidden/locked/interface. The shape KIND cannot change — delete and recreate for that. Patches only the fields you pass. GM-only.
Inputs: drawings (required), sceneIdentifier (required).
update-folderA
Update a sidebar Folder in place — rename it, recolor it, set its raw sort, and/or reparent it (nest under another folder of the same type, or pass parentFolder:"" to move it to the root). Resolves the folder by exact id or exact name+type. Use this to RENAME a folder without the move-documents + delete-folder dance. To reposition a folder, RENAME it — sort does not drive folder order in the v14 sidebar. GM-only.
Inputs: identifier (required), name, sort, type, +2 more.
update-itemA
Update existing world-level Item(s) by id — change name, img, system data, or folder. GM-only. Renaming an UNIDENTIFIED dnd5e item works on the true source name; the echo shows the mystery-mask name plus trueName so the rename is visible.
Inputs: updates (required).
update-journalA
Generic JournalEntry update: rename the entry (name) and/or set page content. Content replaces the target page — pass pageId to target a specific page, newPageName to add a new page, or neither to update the first text page. For quest-style append updates use update-quest-journal instead. GM-only.
Inputs: journalId (required), name, pageId, content, +2 more.
update-lightsA
Edit one or more placed AMBIENT LIGHTS by id (from list-lights): MOVE via x/y, change dim/bright radii, color, alpha, angle, luminosity, attenuation, the animation (animationType/Speed/Intensity — e.g. add torch flicker), the darkness activation range, walls/vision, hidden. Emission fields nest under config internally — patches only the fields you pass, so a partial change never wipes the rest. Unresolved ids reported. GM-only.
Inputs: lights (required), sceneIdentifier (required).
update-noteA
Nudge ONE existing map-note pin by id (the legend→pins review loop): move it (x/y), relabel it, resize/restyle its icon, toggle fog global, or re-point it to a different journal/page. Patches only the fields you pass; at least one is required. Strict scene + note-id resolution. GM-only.
Inputs: noteId (required), sceneIdentifier (required), x, y, +6 more.
update-playlistA
Update a Playlist's document fields: rename, change playback mode, or set the crossfade duration. Does not add/remove tracks. GM-only.
Inputs: identifier (required), fade, mode, name.
update-quest-journalA
Append a new styled section to a quest/journal page from typed blocks (e.g. a heading "Session 3" + paragraphs of what happened) — the §8 session-log/progress path. You supply the words as blocks; the tool styles + appends them. By default appends to the first text page; use pageId to target a page, or newPageName to start a new page. Structuring only.
Inputs: blocks (required), journalId (required), pageId, newPageName, +1 more.
update-regionA
Update ONE region by id: rename, recolor, change visibility, replace its shapes whole, or reshape to a single grid rectangle via the rect convenience (center px + cells + snap — the move/resize you'd do reviewing a teleporter). Patches only what you pass; behaviors are left untouched. GM-only.
Inputs: regionId (required), sceneIdentifier (required), name, rect, +3 more.
update-rolltableA
Update a RollTable's fields (name, description, formula, replacement, displayRoll) and/or its entries, two ways: editResults = TARGETED per-entry edits — name an entry by its roll face (e.g. 7 on a d12) or resultId (from get-rolltable) and patch just its text, linked uuid, weight, and/or range; every OTHER entry (ranges, weights, @UUID item links) stays byte-identical — the right way to fix a typo on one entry of a tuned table. text/uuid REPLACE that entry's content (copy the raw text from get-rolltable and change only what you need). results = DESTRUCTIVE whole-set replace (all entries deleted and recreated with auto-assigned ranges). Bad edits are isolated + reported; an introduced range overlap/gap is warned. GM-only.
Inputs: identifier (required), name, formula, results, +4 more.
update-sceneA
Update an existing Scene document — rename, swap its background image (Data-relative path), toggle navigation, set the navigation label, change dimensions/grid (size/type/distance/units)/padding, token vision, fog mode, lighting (darkness, global light), weather, a nav thumbnail, or the linked playlist/journal ("" clears a link). Also (parity with create-scene) deep-merges a full environment{}/fog{} mood object, re-points the saved camera (initial{x,y,scale}), or re-stamps document flags on an existing scene. Scene-document only: never touches placeables (walls/lights/tokens) and never activates the scene. GM-only.
Inputs: sceneIdentifier (required), fog, name, flags, +22 more.
update-soundsA
Edit one or more placed AMBIENT SOUNDS by id (from list-sounds): MOVE via x/y, resize the audible radius, swap the track (path), change volume/repeat/walls/easing, the darkness activation range, or the listener effects. Patches only the fields you pass. Unresolved ids reported, never fatal. GM-only.
Inputs: sounds (required), sceneIdentifier (required).
update-tilesA
Edit one or more placed TILES by id (from list-tiles). RESIZE via width/height (the tile's on-map size — this is "tile scale"; the top-left corner stays put unless x/y is also given); MOVE via x/y (the new TOP-LEFT in canvas pixels — center-anchor conversion is handled for you); also rotation, alpha, elevation, sort, texture src/tint/fit/scaleX/scaleY (image zoom within the frame), occlusion, light/weather restrictions, video, hidden, locked. Patches only the fields you pass; unresolved ids are reported, not fatal. GM-only.
Inputs: tiles (required), sceneIdentifier (required).
update-tokenA
Edit one or more PLACED tokens on a scene — a token INSTANCE already dropped on the map, NOT the actor's prototype token (that's update-actor). Resolve the scene by id/exact name (default: the ACTIVE scene), then target tokens by tokenIds and/or actorIds (an actor id OR exact name — updates EVERY placed copy of that actor, e.g. all "Dead Guard" corpses). Patch any of: rotation (or randomizeRotation for an independent per-token angle), scale (token art size — sets texture.scaleX/scaleY together), imagePath (RESKIN the placed instance — new token art, still image or animated video, existence-checked so a 404 never breaks working art; no more delete+re-place), elevation, hidden, lockRotation, x/y, name, displayName (nameplate visibility), displayBars (resource-bar visibility), bar1/bar2 (which resource each bar tracks — the health bar is bar1 = attributes.hp), ring (dynamic token ring on/off), and hp (this token's CURRENT hit points, per-token on its own delta — so two copies of one actor can be wounded differently, which update-actor cannot do) — all matched tokens update in one batch. GOTCHA handled for you: a token whose actor had auto-rotate OFF carries lockRotation:true, which HIDES a set rotation — so when you rotate a locked token the tool auto-unlocks it and warns. Reports matched/updated counts + any unresolved ids. GM-only.
Inputs: x, y, hp, bar1, +15 more.
update-userA
Update a Foundry user account: permission role (e.g. demote a trusted player to player), login name, player color, pronouns, or assigned character. GM admin — guarded so the bridge user's own role and the world's last gamemaster cannot be demoted.
Inputs: user (required), name, role, color, +2 more.
update-wallsA
Edit one or more WALLS by id (from list-walls): flip a door to secret (door:2), open/close/LOCK it (ds: 0/1/2), change what it blocks (move/light/sight/sound: 0 none / 10 limited / 20 normal / 30 proximity / 40 distance), set one-way dir, doorSound, or proximity thresholds; MOVE by giving the full segment (all of x0,y0,x1,y1 or c:[4] — a wall never half-moves). Patches only the fields you pass; an off-enum value skips that patch with a warning. GM-only.
Inputs: walls (required), sceneIdentifier (required).
upload-assetA
Plane B (file channel, write). Upload an ASSET (map/token/audio/handout image) from a local file to the Foundry data area over WebDAV and return its public HTTPS URL, so large media bypass the bridge entirely. Missing parent folders are created automatically. ASSETS ONLY — never world-DB files (LevelDB writes while the server runs corrupt it; such paths are refused). PRIVACY: anything under Data/ is served publicly with no auth — do not upload anything sensitive. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: localPath (required), remotePath (required), overwrite.
upload-asset-treeA
Plane B (file channel, write). Recursively upload a LOCAL directory tree of ASSETS to the Foundry data area over WebDAV, preserving the subtree layout (each file → remoteRoot/), creating parent folders as needed. Use for BULK imports — a scene pack's images, a tiles folder — instead of one upload-asset per file. Skips files that already exist unless overwrite:true; optional includeExt filter (e.g. ["webp"]). ASSETS ONLY — refuses live world-DB paths. Reports uploaded/skipped/error counts. PRIVACY: anything under Data/ is served publicly with no auth. Requires MOLTEN_WEBDAV_PASSWORD.
Inputs: localRoot (required), remoteRoot (required), overwrite, includeExt.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 151 tools
The vast majority of tools have clearly distinct purposes, and the detailed descriptions eliminate most ambiguity. However, the sheer number of actor/item creation and editing tools (create-pc, create-pc-from-prefab, create-actor-from-compendium, import-item, add-item) creates potential for misselection despite the clarifying descriptions.
Tool names follow an almost perfectly consistent verb-noun hyphenated pattern (list-, create-, update-, delete-, get-, import-, etc.). A few exceptions like content-audit and asset-info are minor and do not undermine the overall predictability.
With 151 tools, this server is extremely bloated compared to a typical well-scoped MCP server. Even for a comprehensive Foundry VTT integration, the number is far beyond what an agent can efficiently navigate, causing significant selection overhead.
The tool surface is exceptionally comprehensive, offering CRUD for nearly every Foundry document type, plus extensive D&D 5e authoring, compendium, asset, and chat features. Minor gaps exist (e.g., no update-cards tool, no compendium write operations), but they are not critical for the server's core purpose.
Maintenance
Related MCP Connectors
Connect any AI to your Foundry VTT world: actors, combat, dice, journals, tokens, compendiums.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
Model Context Protocol server for the Apideck Unified API. Connect any MCP-compatible agent framework to 100+ accounting systems, HRIS platforms, file storage providers, and more through one integration. More information https://www.apideck.com/mcp-server
Official remote MCP server for Archivist AI TTRPG campaign memory: characters, sessions, and more.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceA Model Context Protocol server that integrates with FoundryVTT, allowing AI assistants to interact with tabletop gaming sessions through natural language to query actors, roll dice, generate content, and manage game worlds.412 npm36MIT
- FlicenseNot gradedqualityNot gradedmaintenanceIntegrates with FoundryVTT tabletop gaming sessions, allowing AI assistants to query game data, roll dice, generate content (NPCs, loot, encounters), manage combat, and provide tactical suggestions through natural language.6 npm-
- AlicenseNot gradedqualityDmaintenanceConnects AI assistants to Dungeons & Dragons 5e game information via the Model Context Protocol, enabling queries for spells, monsters, equipment, and more.54MIT
- AlicenseNot gradedqualityFmaintenanceEnables AI-powered campaign management for Foundry Virtual Tabletop through natural language, supporting multiple RPG systems with tools for quest creation, character management, combat resolution, and more.MIT