Skip to main content
Glama

Tanvo MCP

npm license

The MCP server for Tanvo, an AI image, video and music studio. Ask your assistant for "a renaissance portrait of my cat from this photo", "turn this selfie into a boxed action figure", "a 10-second clip of this product spinning" or "a birthday song for Sam", and it runs it on Tanvo and hands you the file.

Works in Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Cline, Codex, Gemini CLI and any other MCP client. No API key needed to try it: the house image engine runs on the free tier.

What it can do

  • 80 ready-made apps with 350+ preset looks. Pet portraits, figurines, old photo restoration, AI baby, product photos, outfit try-on, trend videos (hotel lobby duet, crying filter, dance), birthday and love songs. Each look is a tuned prompt with a real example output, so you know what you will get.

  • Images on Nano Banana 2 / Pro, Seedream 5, GPT Image 2.5, Qwen Image, Grok Imagine and the free house engine: text to image, edits and multi-photo compositions.

  • Video on Veo 3.1, Kling 3.0, Seedance 2.5, Wan, Minimax and more: text to video, image to video, start and end frames, native sound.

  • Songs on Suno V6: two takes per run, each an MP3 with cover art, a title and the lyrics.

  • Local photos are uploaded for you, and results can be saved straight into a folder.

Related MCP server: PixelDojo MCP

Tools

Tool

What it does

find_apps

Search the apps by what the user wants ("royal pet portrait", "fix an old photo")

get_app

An app's inputs and looks, with example outputs, prompts and cost

generate_from_app

Run an app's look on the user's photos

generate_image

Text to image, or edit / combine photos

generate_video

Text or image to video (returns an id; videos take minutes)

generate_music

Write a song: describe it, bring lyrics, or go instrumental

get_generation

Status and outputs of a run; can wait and save files

list_models

Every model, its options and its credit cost

account

Free tier or account, and the credits left

open_in_tanvo

A link that opens the app, look or studio ready to go

Install

The server runs with npx, so there is nothing to install first (Node 20+).

Claude Code

claude mcp add tanvo -- npx -y tanvo-mcp

Add -e TANVO_API_KEY=sk_... to run on your account.

Claude Desktop, Cursor, Windsurf, Cline, VS Code (the mcpServers block of the client's MCP config)

{
  "mcpServers": {
    "tanvo": {
      "command": "npx",
      "args": ["-y", "tanvo-mcp"],
      "env": { "TANVO_API_KEY": "sk_..." }
    }
  }
}

Gemini CLI

gemini extensions install https://github.com/tanvoai/tanvo-mcp

Codex (~/.codex/config.toml)

[mcp_servers.tanvo]
command = "npx"
args = ["-y", "tanvo-mcp"]
env = { TANVO_API_KEY = "sk_..." }

AI agents that install servers themselves (Cline and others) can follow llms-install.md.

Free tier and API keys

Without a key the server uses Tanvo's anonymous free tier: the house image engine (studio-image-v1), watermarked, a few images per machine per day. Everything else (apps on premium models, video, songs) needs a key. When a call needs one, the server replies with a link that opens the same app or prompt in the browser, where new accounts start with free credits.

Create a key under Settings → API keys. A key runs on your own account: the same models, prices and plan as the website, and every run lands in your history. Each run's price is shown by get_app / list_models, and failed runs are refunded automatically.

Variable

Meaning

TANVO_API_KEY

Your API key (optional)

TANVO_OUTPUT_DIR

Save every result into this folder (optional; each call can also pass save_to)

TANVO_BASE_URL

Another deployment, e.g. a local build (default https://tanvo.ai)

TANVO_ANON_ID

Fix the free-tier id instead of the one stored in ~/.config/tanvo-mcp/anon-id

Examples

Make my dog look like a Renaissance oil painting. ~/Pictures/rex.jpg

find_apps → get_app ai-pet-portrait-generator → generate_from_app with look renaissance and the photo.

Write a birthday song for my sister Ana, she loves surfing and terrible puns. Save it to ~/Music.

generate_music in Describe mode with save_to: "~/Music".

Animate this product photo into a slow 360° spin, 9:16.

find_apps "product spin" → generate_from_app product-spin, then get_generation until it is done.

Privacy

Prompts and photos you pass are sent to tanvo.ai to run the job, under the privacy policy. Content is moderated before anything runs. The server stores nothing except the anonymous free-tier id.

Develop

npm install
npm run build
TANVO_BASE_URL=http://localhost:3210 node dist/index.js

Docs: tanvo.ai/developers/mcp · API reference: tanvo.ai/developers/api · OpenAPI: tanvo.ai/openapi.yaml

Also from Tanvo

License

MIT

Available Tools

10 tools
accountAccount and creditsA
Read-only

Who the server runs as (an API key account or the anonymous free tier) and how many credits are left.

Inputs: no named parameters.

find_appsFind a Tanvo appA
Read-only

Search Tanvo's ready-made AI apps (photo effects, trends, portraits, product shots, video effects, songs) by what the user wants, e.g. 'royal pet portrait', 'action figure of me', 'old photo restoration', 'dancing video', 'birthday song'. Each app has tuned preset looks with real example outputs. Use get_app for the looks, then generate_from_app to run one.

Inputs: kind, limit, query.

generate_from_appRun a Tanvo appA

Run one of Tanvo's apps with one of its looks on the user's photos, e.g. the 'renaissance' look of a pet portrait app. Photos go in the order of the app's inputs (see get_app). Uses the look's tuned prompt and model; details adds the user's own touch. Charges the look model's credits (shown by get_app) and refunds failures. Needs TANVO_API_KEY unless the look runs on the free engine.

Inputs: app (required), look, wait, photos, +2 more.

generate_imageGenerate or edit an imageA

Text to image, or an edit / combination when images are given. Returns the image (inline when small) and its URL. The free tier runs studio-image-v1 without a key; other models need TANVO_API_KEY. Credits are charged per run and refunded if it fails.

Inputs: prompt (required), wait, model, aspect, +4 more.

generate_musicWrite a songA

Makes two takes of a song on Suno V6, each an MP3 with cover art, a title and the lyrics as sung. Modes: Describe (prompt is one sentence about the song; lyrics are written for you), Lyrics (prompt is your lyrics, with [Verse] [Chorus] [Bridge] tags), Instrumental (prompt describes the sound). 60 credits a run; needs TANVO_API_KEY.

Inputs: prompt (required), mode, wait, style, +3 more.

generate_videoGenerate a videoA

Text to video, or animate a still when image is given (optionally ending on end_image). Video takes one to several minutes, so by default this returns at once with an id for get_generation. Needs TANVO_API_KEY.

Inputs: prompt (required), wait, audio, image, +6 more.

get_appShow an app and its looksA
Read-only

One Tanvo app: the photos it needs, its preset looks (each with an example output and the prompt behind it), the model it runs on and the link that opens it in the browser.

Inputs: app (required).

get_generationCheck a generationB
Read-only

Status and outputs of a run started earlier (poll this for videos).

Inputs: id (required), wait, save_to.

list_modelsList modelsA
Read-only

Every image, video and music model with the options each accepts (aspect, resolution, duration, format, audio), how many reference images it takes, whether the free tier covers it, and its credit cost at default settings.

Inputs: kind.

open_in_tanvoOpen in the browserA
Read-only

A link that opens Tanvo ready to go: an app on a chosen look, or the image / video / music studio with a model and prompt filled in. Use it when the user wants to upload photos and tweak settings themselves, or has no API key.

Inputs: app, kind, look, model, +1 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 10 tools

    Disambiguation4/5

    The generation tools are cleanly separated by media type (image, video, music) and by source (generate_from_app vs generate_image), and discovery tools (find_apps, get_app, list_models) have distinct roles. The only mild tension is generate_image vs generate_from_app (both yield images) and open_in_tanvo vs the generate_* family (both can set up a creation), but descriptions make the boundaries clear.

    Naming Consistency4/5

    Eight of ten tools follow a consistent snake_case verb_noun pattern (generate_image, generate_video, generate_music, find_apps, get_app, get_generation, list_models, generate_from_app). Deviations are minor: 'account' is a bare noun, and 'open_in_tanvo' is verb_preposition_noun, both still readable and unambiguous.

    Tool Count5/5

    Ten tools is well-scoped for a media-generation service: discovery (2), details (1), generation (4), status (1), account (1), and a browser escape hatch (1). Each tool earns its place with no redundancy.

    Completeness4/5

    The surface covers the full create-and-poll lifecycle across image, video, music, and preset apps, plus model catalog and account/credit visibility. Minor gaps remain: no cancel/delete for in-flight generations and no credit top-up or history listing, though agents can work around these.

    Maintenance

    ActivitySlowing
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers