LUVI Creator

npx add-mcp 'https://www.luvicreator.com/api/mcp'

72 LUVI tools inside Claude and ChatGPT

LUVI is an MCP server. Connect it once and your assistant can pick a model, quote the cost, generate, file the result and read it back, all against your own LUVI account and credits. Each tool below is explained in plain words.

Claude (claude.ai)

1. 1Open Customize → Connectors, press + Add and choose Add custom connector.
2. 2Name it LUVI and paste the endpoint below.
3. 3Sign in with your LUVI account when Claude asks, and approve the connection.
4. 4Start a chat and ask for a video; Claude quotes the cost before it spends credits.

ChatGPT

1. 1Open Plugins, press + and choose Add custom MCP server. Name it LUVI and paste the endpoint below.
2. 2Set authentication to OAuth, tick “I understand and want to continue” and press Create as a plugin.
3. 3Sign in with your LUVI account when ChatGPT asks, and approve.
4. 4In a chat, type @, pick LUVI and ask for what you need. Generating needs ChatGPT Business or Enterprise/Edu on the web; on Pro, ChatGPT can only read.

Endpoint https://www.luvicreator.com/api/mcp ·OAuth sign-in with your LUVI account; nothing is shared with the assistant's provider.

Step-by-step guides

• How do I connect LUVI to Claude?
• How do I connect LUVI to ChatGPT?
• How do I connect LUVI to Claude Code?
• How do I upload a file to LUVI from Claude or ChatGPT?
• How do I build a workflow from Claude?
• How do I disconnect Claude or ChatGPT from LUVI?

Generate

Generate

generate

Starts a video, image, audio or 3D generation and spends your credits; LUVI fills the defaults, queues it if needed and moves settings in the prompt ("9:16", "4K") into the parameters. It stops without charging until the assistant has read the model's LUVI Prompt Engine rules.

Quote the cost

read-only

estimate_cost

Works out how many credits a generation will cost, with the same pricing the charge uses, so you see the price before anything is spent.

Get the result

read-only

get_generation_status

Follows a generation from its place in the queue until it is done, then shows images and plays audio in the chat, gives shareable links and reports the credits it actually cost. For a layer split, it also says which file is the base image and which element each layer holds.

Cancel a queued generation

cancel_generation

Removes a generation that is still waiting in line, so it never starts and costs nothing. One that has already started cannot be stopped and is charged when it finishes; the tool says so.

Read a model's exact settings

read-only

get_model_schema

Lists every setting a model accepts, with its allowed values and defaults, so the assistant sets resolution, duration or aspect ratio correctly instead of guessing.

Browse the models

read-only

list_models

Lists every active LUVI model for video, image, audio and 3D, so the assistant can pick the right one for your job.

LUVI Prompt Engine

Read the LUVI Prompt Engine rules

read-only

get_prompt_engine_rules

Gives the assistant the LUVI Prompt Engine rules for the chosen model, so it writes the prompt the way that model understands best. They end with a receipt showing the rules were read, which the assistant needs before it can send a prompt for that model.

Read the LUVI Prompt Engine rules (older name)

read-only

get_prompt_manual

The same LUVI Prompt Engine rules under the tool's older name, kept for assistants that saved their tool list before the rename. It returns exactly what “Read the LUVI Prompt Engine rules” returns.

Check the prompt

read-only

check_prompt

Checks a drafted prompt against the LUVI Prompt Engine rules a machine can check (syntax, ratios or resolutions in the text, length) and returns a corrected version. It lists the rules it checked; the writing advice stays with you.

Workflows

Create a workflow

create_workflow

Starts a workflow — empty, or from a template — as a chain of models, prompts and references you can review on the canvas before it runs. From a template the whole chain arrives in one step, and the assistant adapts it to your subject and your files.

See how models connect

read-only

get_workflow_ports

Tells the assistant which inputs each model and each built-in node of a workflow takes, so it wires them correctly.

Build the workflow

wire_workflow

Adds nodes and wires or a whole LUVI template. A wrong wire saves nothing; a blank field saves. Deletes steps only if the owner allows it, and takes a model's prompt only after the assistant has read its LUVI Prompt Engine rules.

Quote a workflow

read-only

estimate_workflow

Shows what a whole workflow will cost in credits, node by node, next to your balance. A ~ marks a number that can still move — for example media another node has not made yet.

Run a workflow

run_workflow

Starts the workflow and spends credits, but only if you allowed the assistant to run it on the canvas with a credit ceiling and an expiry; otherwise it hands you a link. If a prompt breaks its model's LUVI Prompt Engine rules, it stops once before any charge.

Read a workflow

read-only

get_workflow

Shows a workflow's nodes and connections, and anything that would stop it from running.

Follow a run

read-only

get_workflow_run

Shows the progress and results of every step in a workflow run, and the credits charged so far.

List your workflows

read-only

list_workflows

Shows your workflows, most recently edited first, with a link to each.

Copy a workflow

copy_workflow

Copies one of your own workflows into a new one, with every node, setting and cable, so the assistant only changes what differs. The original stays as it is, and nothing runs.

Tag a workflow

tag_workflow

Sets word tags on one of your workflows, or tags it to other projects so it also shows there. The workflow itself does not change.

Stop a workflow run

cancel_workflow_run

Stops a run that is still going. Steps that have not started are neither run nor charged; generations already sent finish and are charged.

Retry a failed workflow step

retry_workflow_step

Retries one failed step of a run the assistant started, spending credits only for that step. It works only while you have the workflow armed on the canvas; otherwise you get a link to retry it yourself.

Library and projects

Browse your library

read-only

list_library

Lists everything you have generated and uploaded, newest first, filtered by project, folder or media type, so the assistant can find earlier work and build on it.

Open one item

read-only

get_asset

Shows a single item from your library with a share link, what it cost and a reference the assistant can use in the next generation.

Search your library

read-only

search

Finds your generations and uploads by words in their prompt, their file name or the name you gave them. ChatGPT uses it to look through your work.

Open a search result

read-only

fetch

Returns the full details of one search result: prompt, model, media type, size and cost.

Edit library items

update_assets

Favourites, rates, tags or renames results and uploads in your library, just as the Library page does. Nothing is deleted.

Move library items to the trash

trash_assets

Moves your own results or uploads to the Library's trash. For 24 hours they can be brought back; after that they are deleted for good, as on the website.

Restore library items

restore_assets

Brings results or uploads back from the Library's trash within 24 hours of trashing them.

Make a frame-rate copy of a video

make_video_fps_copy

When a model refuses your video only over its frame rate, makes a copy at a rate the model takes, after you say yes. The copy is a new file in your library; your original stays as it is. No credits, but it counts toward storage.

List a video's markers

read-only

list_video_markers

Shows the markers on one of your videos: the notes pinned to moments on its timeline in the LUVI Player.

Add a marker to a video

add_video_marker

Pins a note to a moment on the timeline of one of your videos, where it shows in the LUVI Player. The video itself is not changed.

Remove a marker from a video

remove_video_marker

Takes one marker off one of your videos. Only the note goes; the video stays as it is, and the marker can be added again.

List your projects

read-only

list_projects

Shows your projects and the ones shared with you. When you name none, results go to a project created for the connection.

Create a project

create_project

Creates a new project in your account so the next results land in it.

Rename, bookmark or rate a project

update_project

Renames, bookmarks or rates one of your own projects, as the Projects page does.

Move a project to the trash

trash_project

Moves one of your own projects to the trash with everything in it. For 24 hours it can be brought back; after that it is deleted for good. A locked project is refused.

Restore a project

restore_project

Brings one of your projects back from the trash, with everything in it, within 24 hours of trashing it.

List folders

read-only

list_folders

Shows the folders inside a project.

Create a folder

create_folder

Creates a folder in a project, inside another folder if you like, so new results can be filed straight into it.

Rename a folder

rename_folder

Changes a folder's name; everything inside it and every existing reference stays the same.

Move items into a folder

move_to_folder

Sorts results and uploads that are already in your library into a folder of the same project, or back to the project root.

Move a folder to the trash

delete_folder

Puts a folder and the folders inside it in the trash for 24 hours. The files are not deleted; they stay in your library at the project root.

Restore a folder

restore_folder

Brings a trashed folder back within 24 hours, with its files where they were.

Upload

Upload a file directly

create_upload_ticket

When Claude can run code, it uploads a file you attached to the chat straight into your project, with no browser step: images up to 50 MB, other files up to 1 GB.

Get an upload link

get_upload_link

Gives you a link that opens LUVI's upload window in your browser, already set to the right project and folder, for a file on your computer or attached to the chat.

Check upload limits

read-only

get_upload_limits

Tells the assistant the maximum file sizes and which formats are accepted or converted, before it tries an upload.

Account

See which account is connected

read-only

get_account

Shows the LUVI account the assistant is signed in as and the last generation request LUVI received from this connection, so you can confirm the account or check whether a request that seemed stuck ever arrived.

Check your balance

read-only

get_credit_balance

Shows your credit balance and how much this connection can still spend under its cap.

See what you spent

read-only

list_transactions

Lists your credit history as it appears under Account → Billing: every charge, purchase and refund, with the prompt and model behind each generation.

Check your storage

read-only

get_storage_usage

Shows how much of your storage is used, by kind of file, and whether new uploads and generations are blocked because it is full. Same figures as Account → Storage.

Read this connection's rules

read-only

get_connection_rules

Shows the rules you set for this connection — banned models, a credit ceiling per generation, a fixed resolution or aspect ratio — so the assistant plans within them.

Add a rule to this connection

set_connection_rule

Saves a rule you ask for, such as “never use Veo” or “nothing above 500 credits”, and applies it to every later generation and workflow run. It can only tighten; you loosen or remove rules in Account → Integrations.

Star a model

favorite_model

Stars or un-stars a model for you, like the star on the Models page. Starred models come first in your model lists.

List your model sets

read-only

list_model_sets

Shows your model sets (your own named groups of models from the Models page) and your starred models.

Create or change a model set

save_model_set

Creates a model set, or renames one and adds or removes models in it. A set is never deleted from here; you do that on the Models page.

Read the help center

read-only

read_help_article

Finds and reads LUVI help-center articles, so the assistant answers “how do I…” questions from LUVI's own guides.

Other

List workflow templates

read-only

list_flow_templates

Shows the ready-made workflows you can start from: LUVI's own, each tested end to end and with the credits one full run cost, plus your own and your team's. For each one, what it makes and which files it needs.

Find styles and poses

read-only

search_flow_catalog

Searches the Prompt Lab (style, film stock, palette, lens, camera move, light) and the Pose Atlas, so the assistant puts a real style or pose on every node of a workflow instead of guessing.

List your video edits

read-only

list_edits

Lists your own video edits, most recently changed first, with their project, aspect ratio, length and a link to open each one. A teammate's edit is not listed.

Read a video edit

read-only

get_edit

Shows one of your edits as it is saved: its tracks, every clip with its place in time and the file it uses, the text of its text and caption clips, and its markers. The assistant reads this to work on the edit.

Follow an export, separation or captions job

read-only

get_edit_job

Shows where an export, an audio separation or a captions job stands, and once it is done, the files it added to your library.

Create a video edit

create_edit

Starts a new video edit from files in your own library, placed one after another as "Send to editor" does. It is free and changes no other edit.

Add to the end of an edit

append_to_edit

Adds files from your library and text cards to the end of one of your edits. It never trims, moves or deletes anything, and the edit as it was is saved first, so you can restore it from the edit's version history.

Export a video edit

export_edit

Renders one of your edits to an MP4 in your library. 480p and 720p are free. 1080p, 4K where offered and upscales up to 8K cost credits: you see the price first (for an upscale, an estimate) and it starts after your yes. It can also balance the sound or export part of the edit.

Save a frame from an edit

save_edit_frame

Saves the picture at a chosen moment of one of your edits to your library as an image. It is the video clip's own frame, without the text and graphics drawn over it. Free; the image counts towards your storage.

Split vocals & instruments

separate_audio

Splits the sound of one of your files into four tracks — vocals, drums, bass and the rest — each a new audio file in your library that counts towards your storage. Paid: the assistant shows the price and starts only after your yes. Separate only sound you have the rights to.

Create captions for an edit

create_captions

Turns the speech in one of your edits, or part of it, into captions. Paid: the assistant shows the price and starts only after your yes. You place the captions yourself from the Captions window. Its sound mix is a new file that counts towards your storage.

Change a video edit

edit_timeline

Changes one of your edits: trims, splits, rolls, joins, moves, duplicates or removes clips; sets transitions, speed, texts, blend, stabilising, markers, tracks, frame rate, export range; grades colour; mixes sound (faders, buses, EQ, dynamics); keyframes. All or nothing.

Look inside your videos

read-only

inspect_clips

Before cutting, the assistant looks at up to 10 of your videos, or only the parts an edit uses: length, sound, cuts and jumps, black, frozen or silent parts, where the sound's hits and beats land, and up to 16 frames; music and sound clips too. Free; a new video takes 10–25 s.

Go back to a saved version of an edit

restore_edit_version

Shows the versions of one of your edits that were saved before each assistant change, and brings back the one you choose when you ask to go back. The edit as it is now is saved first, so going back can be undone too.

List your colour looks (LUTs)

read-only

list_luts

Lists the LUT files you uploaded, your colour looks, by name, so the assistant can put one on clips in your edit.

Write down what is said in your videos

transcribe_clips

Writes down the speech in up to 10 of your videos, so the assistant can cut by what is said. Paid: the assistant shows the price and starts only after your yes. The sound is sent to a speech recognition service; a video written down before is not charged again.

Which models can it use?

Every model in the public catalogue, with the same prices and parameters. Browse the model catalogue →

Questions

What can Claude or ChatGPT do with LUVI?

Pick a model, read its exact settings, quote the cost in credits, generate video, image, audio and 3D, build and run multi-step workflows, put together video edits and export them, bring your own files in, and read back your library and spending. That is 72 tools, all acting on your own LUVI account.

Does connecting LUVI spend credits?

No. Credits are spent only when a generation or a workflow run starts, and the assistant can quote the price first with estimate_cost. Every charge appears under Account → Billing, like any other.

Which assistants work with it?

claude.ai on the web and desktop, Claude Code, and ChatGPT with developer mode. Claude Code signs in with an access token you create under Account → Integrations; any other MCP client that sends that token works the same way.

Can the assistant use a file from my computer?

Yes. It can give you an upload link that opens LUVI's upload window in your browser, or, when Claude can run code, upload a file you attached to the chat directly. Images can be up to 50 MB and other files up to 1 GB.

How do I limit what the assistant can spend?

Under Account → Integrations you can set a credit cap for each connection; once it is reached, that connection stops generating. Revoke removes the connection at once, and nothing it already made is deleted.

Can the assistant change my password or payment details?

No. It signs in through LUVI's own page and works only through these 72 tools, and none of them can change your password, payment methods or team.