MCP: Gomo Sounds inside your AI assistant

How much of the game already has a sound (what ships to the game), in % and per tab, and every missing event with its id. The assistant calls it at the end and asks whether to fill the rest.

Searches entities by name, game, event or description (accent-insensitive) and returns everything about each: image, events, every variation with id, prompt used, duration, cost, whether it's chosen and the audio URL. Also finds folder events (music, ambience, interface) by name, description or folder.

An entity's full profile by id.

The account's games with how many entities, events and sounds each has, the credit balance, the price per second of sounds (creditsPerSecond) and the price per minute of music per quality level, with your plan's maximum length.

A game's entities, with how many events already have a chosen sound.

The voice library for character lines: each voice with its name, who it is (man, woman, boy, girl), character or narrator style, a description, the registers it has (loose, conversational, read) and samples to listen to. Also the emotions, the languages (Portuguese and English), the longest line and the price.

The game's folders and subfolders, with the events the entities inside inherit: duration, loop and, for a track, what is saved in Compose (mode, style, tempo, key, parts).

The assistant's full guide: every rule (how games, folders, sounds and versions work (and the label of each thing of the game), duration and cost, the yes before spending, Godot, links). The assistant reads it at the start of every conversation, since some apps cut the server instructions short.

Creates a game. It is the first step: it comes with the Ambience, Sound effects and Interface folders. sound_direction is the style every generation follows; template adds genre folders (strategy, platformer, board, puzzle).

Creates several tracks at once in the Music folder (or a subfolder of it), free, without generating: name, prompt, duration, loop and the screen form (style, parts, tempo, key, time signature, voice), shown on the site right away; the style and the voice go into the prompt as words. A track that already exists is updated with what is sent. Returns each track's page and the cost of generating each one and the batch, per quality level.

Creates ONE voice line (what a character says) in a game, in the Voices folder (a new game is born with it; an older one gets it with the first line), named in the game's language. Takes the name, the exact text, the language (the game's by default), the voice, the register and the emotion. Free: nothing is generated yet.

Creates an entity: a thing in the game that makes noise (unit, building, block, object, button, screen). Actions like Undo or Level complete are not entities, they are events. Inside a folder, it inherits the folder's events.

Creates an event on an entity (entity_id) or a folder (folder_id). The name is what happens (Move, Attack, Click), never "Sound": the game calls the sound by that name. Actions with no owner go in the Interface folder. loop is required: true for what repeats while something keeps happening (ambience, music, engine, fire and movements like footsteps, running and wings; footsteps are a 6-10 s cycle), false for a single action (hit, click, cry), at least 2 s. The server refuses one-shots under 2 s and loops under 6 s.

Deletes a sound or a track for good, with its versions, like the site's Delete button. Only when you ask, and the name must match (a mixed-up id never deletes the wrong sound). A favorite version or one chosen in another sound stays; with a generation running, it waits for it to finish.

The sound direction of the game (game_id) or of a folder (folder_id): the style every generation there follows.

Creates a folder in the game, in any tab (section: ambience, effects, interface, music) or inside another folder, optionally already holding entities. Ambience by place, interface by screen, effects by kind of thing.

Moves entities into a folder. Without folder_id, removes them from their folder.

public shows in the Explore gallery; private is owner-only (paid plan required).

The same assistant as the site's "I want maximum refinement!": from the name and "what the game is like" (genre, world, tone, sound style, camera, what is in the game) come the description, the sound direction and the folders of each tab with their events. Nothing is created: you review and create_game creates.

One sound per entity in a folder (e.g. Footsteps for each unit). '{entity}' and '{description}' in the prompt become each one's name and description. loop is required, as in create_event. Without max_credits it only shows the list and the cost; with max_credits equal to the cost, it creates the events and generates. An event with no chosen sound gets all the new variations chosen. Accepts channels, like generate_sounds.

Generates variations for the event. Costs credits: seconds × price per second × variations. Without a prompt, uses the event's description. If the event has no chosen sound yet, all the new variations become chosen and the game picks one at random (auto_choose=false turns it off); a choice you already made never changes. channels: auto (mono for what the game positions, stereo for Ambience backgrounds and music), mono or stereo; it is saved on the event.

Generates ONE version of a track for an event of the Music tab: the prompt goes to the model exactly as written (English comma-separated tags: genre, mood, instruments, tempo), duration, loop, quality level (draft to max, with a price per minute) and instrumental or with lyrics. Also the Compose mode (parts, tempo, key, time signature, style, voice); the style and the voice become words in the prompt itself. Waits up to 10 minutes; with wait=false it only queues and returns the id for job_status (for several tracks: the account generates one at a time and the rest wait their turn). Before the first track, the user accepts the rights notice on the site.

A new version from an existing one: regenerates only a section (repaint, with start and end in seconds), extends (extend, with the seconds to add) or makes a similar track (cover, with the closeness). Charges the generated seconds at the source version's level; lyrics, tempo and key come from the source, and the note is added to the end of its prompt as written.

Where the generations are: the ids from generate_music/derive_music with wait=false (or, without ids, every unfinished one of the account). Each one is waiting its turn, in the engine queue, generating, done (with the new versions, chosen if the sound had none) or failed (the reason; a failure is not charged).

Generates 1 to 4 versions of a voice line. Costs credits: the line's estimated seconds (characters ÷ characters per second, at least 1 s) × price per second × versions, with the same cost approval as generate_sounds. The versions are never chosen automatically: you listen to them on the site and pick the ones that go to the game.

Marks the chosen variation. mode: add (adds to the chosen ones), remove (removes), only (just this one).

Uses the same variation in other events (event_ids), or in every event with the same name (same_name), without generating again.

Records your yes to the cost: it lasts until that total is spent, for up to 6 hours (a whole batch of tracks fits in one yes). The assistant shows the plan with the total ("May I generate this for 63 credits?") and only calls it after your answer; without it, generating is refused. In the plugin, the Generate button does not need it.

Returns a link to download an audio file from YOUR account (the audio URL from find_sounds or get_entity). The link opens in any browser for 24 hours, without signing in. Explore sounds belong to other people and don't go through MCP: download them on the website (free, with an account) or generate your own.

Returns the link to the .zip with the game's chosen sounds and the manifest.json. The link opens in any browser for 24 hours.

For engine plugins: the game's chosen sounds in the neutral format (manifest.json), each with a download link (24 hours) and the suggested path in the project.

For engine plugins: sends the project's scenes (up to 20 at a time) to the game's Inbox. A scene that is already an entity only gets its image refreshed; sending again never duplicates.

For engine plugins: tells the site the chosen sounds were just installed in the project.

The kit to put the Godot plugin into the project: current version, whether the connected plugin is up to date (the assistant checks before asking you to sync or install), .zip link, folder, whether the plan allows connecting and the steps. The assistant downloads and extracts it (with your ok, deleting nothing); you enable the plugin, connect and click Sync.

What the plugin sent from the project: when the last sync was, the scenes waiting in the Inbox (with the tab suggested by their path) and the ones that already are entities. The assistant checks here that the sync arrived.

Puts Inbox scenes in a tab, like dragging on the site: automatic (by the scene path in the project) or into a chosen tab and folder. Each scene becomes an entity ready for events.

Accepts the current Terms of Use and Privacy Policy, like the button on the site. When the Terms change, the assistant asks whether you accept (with the links) and only calls it after your yes.