Mixer AI

io.github.y110692Published by GitHub account y110692v1.0.0Updated Oct 30, 2026

Create and edit images, video, music and voice with 100+ AI models and Canvas.

VerifiedStreamable HTTPWeb executableMedia & DesignAI & ML

Overview

AI-generated overview

Remote MCP server that lets an assistant generate and edit images, video, music and voice with 100+ AI models and a Canvas workspace.

What you can do
  • Create and edit photos, video, music and voice-overs from prompts or attached files
  • Browse live models and options, and estimate or quote coin prices before running a job
  • Submit generations, track status and retrieve results, and search earlier work
  • Upload attachments and attach them to Canvas nodes
  • Read, preview and apply changes on a Canvas, and open an interactive Canvas in MCP Apps clients
When to use it
Useful when you want media generation and editing driven from a chat client such as Claude, ChatGPT, Cursor, Codex or VS Code, including storyboarding on a Canvas, without managing API keys yourself.
Requirements
A remote Streamable HTTP endpoint with OAuth 2.1 (PKCE) sign-in to a Mixer AI account; no API keys or local runtime. A paid Mixer AI coin balance is needed for generations. Some features need a client that supports MCP Apps or a JavaScript-capable browser.
Before you install
Generations spend Mixer AI coins from your account; the assistant is expected to show the exact model and price first, and only an explicit approval counts as consent. Uploads are limited to 30 MiB for photos and 100 MiB for video and audio, and reference files are kept for one day (50 days on a canvas). Connecting grants the assistant access to your account until you disconnect it.

Installation

In SourceWeft

  1. Open Mixer AI in the dashboard and add it to a workspace.
  2. Connect the required credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.y110692%2Fmixerai-mcp/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.

{
  "mcpServers": {
    "mixer-ai": {
      "type": "http",
      "url": "https://mcp.mixerai.org/mcp"
    }
  }
}

README

[Mixer AI]

Mixer AI MCP

Create and edit photos, video, music and voice with 100+ AI models — right inside ChatGPT, Claude, Claude Code, Cursor and Codex.
Setup guide · mixerai.org · На русском


Server URL: https://mcp.mixerai.org/mcp Transport: Streamable HTTP · Auth: OAuth 2.1 with PKCE (sign in with your Mixer AI account — no API keys)

What you can do

TaskExample prompt
Create a photo@Mixer AI award winning portrait photograph of a bizarro fashionista sphere cyborg, pastel yellow clothing, bright yellow background, minimalism
Edit a photo@Mixer AI remove the helmet and change her hair to light brown (attach the photo)
Create a video@Mixer AI a showdown on the dusty main street of an Old West town, slow push-in, 4 seconds
Edit a video@Mixer AI replace the office with a space station, keep the motion and the hero (attach the video)
Make music@Mixer AI create a 20-second instrumental for a travel reel
Voice over@Mixer AI narrate this short script in a calm voice
Work on the canvas@Mixer AI build a 4-scene storyboard on my canvas and animate each scene

[Create a photo] [Edit a photo]

Models include Nano Banana, GPT Image, Seedream, Seedance, Kling, Veo, Grok Imagine and more — the assistant picks the right one or uses the one you name.

Connect

Claude (claude.ai and Claude Desktop)

  1. Open Add custom connector, then enter the connector name Mixer AI and URL https://mcp.mixerai.org/mcp. Do not rely on the link to prefill the form.
  2. Click Add, sign in to Mixer AI through OAuth and allow access.
  3. Enable Mixer AI in your Claude conversation and ask for a supported task. If a file upload from Claude's code execution gets HTTP 403, check Settings → Capabilities → Code execution and file creation → Allow network egress, or use the browser upload page.

ChatGPT

  1. Open chatgpt.com/plugins, click + → Add custom MCP server (available on eligible plans).
  2. Name: Mixer AI, URL: https://mcp.mixerai.org/mcp, authentication: OAuth → Create.
  3. Sign in to Mixer AI and ask in a chat: @Mixer AI …

Claude Code

bash
claude mcp add --transport http --scope user mixerai https://mcp.mixerai.org/mcpclaude mcp login mixerai

Then verify the mixerai connection using /mcp inside Claude Code.

Cursor

Install in Cursor — or add to ~/.cursor/mcp.json:

json
{ "mcpServers": { "mixerai": { "type": "http", "url": "https://mcp.mixerai.org/mcp" } } }

The repository also includes a Cursor Marketplace plugin package with .cursor-plugin/plugin.json, root mcp.json, and a compact skills/mixerai/SKILL.md. The plugin connects to the same remote OAuth server; no bundled binaries, API keys or local generation services are needed. Cursor discovers this native plugin format and its relative logo at assets/logo.png. The Marketplace submission requires a separate owner action at cursor.com/marketplace/publish. The LICENSE file proposes MIT for the public integration files; the rights holder must approve the license and copyright notice before distribution. The Mixer AI brand name and logo are not licensed as trademarks by MIT.

Codex

bash
codex mcp add mixerai --url https://mcp.mixerai.org/mcpcodex mcp login mixerai

VS Code

Add a remote MCP server to .vscode/mcp.json, then authorize it through your VS Code MCP panel:

json
{ "servers": { "mixerai": { "type": "http", "url": "https://mcp.mixerai.org/mcp" } } }

Any other MCP client

Add a remote MCP server under the alias mixerai with URL https://mcp.mixerai.org/mcp (Streamable HTTP) and sign in with OAuth. If your client does not support MCP Apps, tools work in text mode and can return links to open in your browser.

Using the tools

  • Check the connection with aimixer_profile and find live image, video or audio models with aimixer_catalog and aimixer_options.
  • Price without charge via aimixer_estimate; then use aimixer_quote → your explicit approval → aimixer_submit → aimixer_status → aimixer_result. Recover older work through aimixer_history_search.
  • For an attachment, send its actual file bytes using supported ChatGPT fileParams, aimixer_chatgpt_upload, or aimixer_media_upload_prepare. Otherwise use the MCP App file picker or the temporary web page from aimixer_media_upload_session. An attached chat image alone is not a remote file. Ordinary uploads return a reference you then attach to Canvas; aimixer_canvas_import_attachment can directly attach to an existing node or create a media node. Neither operation starts paid generation.
  • For Canvas use aimixer_canvas_capabilities, aimixer_canvas_read, aimixer_canvas_preview / aimixer_canvas_apply. Open an interactive Canvas only in clients supporting MCP Apps (aimixer_canvas_open); other clients can use aimixer_canvas_view_session in a JavaScript-capable browser. Price and approve Canvas execution separately.

The authoritative tool names and input schemas are in lhm.plugin.json. The MCP server shows only tools allowed for the connected account and host; optional UI tools may be absent.

Pricing and consent

  • Connecting is free. Generations are paid in Mixer AI coins from your account.
  • The assistant must show the exact model and coin price first. For a single generation, approve through the MCP App button, the authenticated website, or a literal unambiguous answer such as “I agree” / «Согласен» to that price question. A bare "yes", generic chat acknowledgement or OAuth login is not payment consent. For batch and Canvas execution follow the separate App/website confirmation.
  • Attach your own files: photos up to 30 MiB, video and audio up to 100 MiB, subject to each model's format and duration restrictions.
  • Reference files for a generation are kept for one day; files added to a canvas — 50 days.
  • If an OAuth login returns to the wrong application, reconnect using the mixerai entry and complete sign-in from that client. If tools are stale in ChatGPT after a server release, select Refresh tools / «Обновить инструменты». For an untransferred file, use the upload window or browser page.
  • Disconnect an assistant anytime in your Mixer AI profile → Connected apps.

Links


На русском

Mixer AI MCP — генерация и редактирование фото, видео, музыки и озвучки на 100+ ИИ-моделях прямо в ChatGPT, Claude, Claude Code, Cursor и Codex.

Адрес сервера: https://mcp.mixerai.org/mcp · вход через аккаунт Mixer AI, без ключей.

  • Claude: откройте окно добавления коннектора, укажите имя Mixer AI и адрес сервера, затем войдите через OAuth. Автозаполнение полей по ссылке не гарантировано. При блокировке загрузки проверьте Settings → Capabilities → Allow network egress.
  • ChatGPT: chatgpt.com/plugins → «+» → добавить свой MCP-сервер с этим адресом, вход через OAuth.
  • Claude Code: claude mcp add --transport http --scope user mixerai https://mcp.mixerai.org/mcp, затем claude mcp login mixerai и проверка /mcp.
  • Cursor: установить в Cursor.
  • Codex: codex mcp add mixerai --url https://mcp.mixerai.org/mcp, затем codex mcp login mixerai.
  • VS Code: .vscode/mcp.json: {"servers":{"mixerai":{"type":"http","url":"https://mcp.mixerai.org/mcp"}}}, затем вход через OAuth.

Подключение бесплатно. Инструменты работают и в текстовом режиме, если клиент не поддерживает MCP Apps. Доступны фото, видео, музыка, озвучка, канвас и передача вложений, когда это разрешено каталогом моделей и возможностями клиента. Вложения нужно передавать реальными байтами (либо через окно загрузки), они не становятся доступными MCP автоматически. Фото до 30 МиБ, видео/аудио до 100 МиБ. Обычные референсы хранятся до одного дня, прикреплённые к канвасу медиа до 50 дней.

Для одиночной платной генерации ассистент сначала показывает модель и точную цену. Монеты списываются только после подтверждения кнопкой в окне, на сайте или буквальным ответом «Согласен» на вопрос о конкретной цене. Просто «да» недостаточно. Для пакетов и запуска канваса требуется отдельное подтверждение в приложении или на сайте. При ошибке входа переподключите mixerai в исходном клиенте; после релиза новых инструментов в ChatGPT нажмите «Обновить инструменты».

Подробная инструкция: https://mixerai.org/knowledge/connect-mcp/

Source: README.md at commit cb840e0

Tools

0
This server asks you to sign in, so SourceWeft could not read its tools without an account and has not indexed them. What a connected account sees can depend on that account and its permissions.

Version history

1
  1. v1.0.0LatestOct 29, 2026

Score

85% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts