> ## Documentation Index
> Fetch the complete documentation index at: https://docs.magnific.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Magnific MCP

Magnific MCP is a remote server built on the [Model Context Protocol](https://modelcontextprotocol.io). Once you connect it to an AI assistant, your agent can generate images and video, train consistent characters, upscale assets, and browse your generation history straight from the chat. Everything runs on your Magnific account and uses your existing credits.

You don't need to manage an API key. The first time a client connects, you sign in to your Magnific account and that's it.

## Getting started

<div className="my-11">
  <Columns>
    <Card title="1. Have a Magnific account" icon="user" href="https://magnific.com" cta="Sign in">
      Any Magnific account works. MCP tools share the same credit balance as the in-app product.
    </Card>

    <Card title="2. Add the server" icon="plug">
      Add `https://mcp.magnific.com` in your AI client's MCP settings.
    </Card>

    <Card title="3. Approve the OAuth prompt" icon="shield-check">
      Sign in to Magnific in your browser to grant the client access.
    </Card>

    <Card title="4. Start creating" icon="rocket">
      Ask your assistant to generate, upscale, or browse. It will pick the right Magnific tool on its own.
    </Card>
  </Columns>
</div>

## Server endpoint

The Magnific MCP server lives at:

```text theme={null}
https://mcp.magnific.com
```

It uses the **streamable HTTP** MCP transport, so any modern MCP client can connect directly. You don't need to install a bridge or run anything locally.

## Authentication

Magnific MCP uses **OAuth 2.0** with your Magnific account. The first time a client connects, it opens a browser window so you can sign in and approve access. After that, the client keeps the session and won't ask you again.

### Discovery endpoints

MCP clients discover the OAuth configuration through standard well-known URLs:

| Resource                      | URL                                                               |
| ----------------------------- | ----------------------------------------------------------------- |
| Protected resource metadata   | `https://mcp.magnific.com/.well-known/oauth-protected-resource`   |
| Authorization server metadata | `https://mcp.magnific.com/.well-known/oauth-authorization-server` |

Most clients pick these up automatically when they hit the MCP endpoint without a valid token, so you usually don't have to touch them.

## Connect your client

Every client asks for the same handful of details. Keep these on hand before you start:

| Field              | Value                                                    |
| ------------------ | -------------------------------------------------------- |
| **Name**           | `Magnific`                                               |
| **URL**            | `https://mcp.magnific.com`                               |
| **Authentication** | OAuth (sign in with your Magnific account when prompted) |

<Tabs>
  <Tab title="Claude Web">
    1. Open Claude and click your profile menu, then choose **Customize**.
    2. Go to **Connectors** and click **Add connector**.
    3. Pick **Add custom connector**.
    4. Enter the **Name** (`Magnific`) and **URL** (`https://mcp.magnific.com`).
    5. Finish the **OAuth** sign-in to Magnific when Claude opens it.
    6. **Start chatting.** Ask Claude to generate, upscale, or look through your creations and it will reach for the right Magnific tool on its own.
  </Tab>

  <Tab title="ChatGPT">
    1. Open ChatGPT and click the **settings icon** in the top-right corner.
    2. Go to **Advanced settings** and turn **Developer mode** on.
    3. Open **Apps** and click **Create app**.
    4. Enter the **Name** (`Magnific`) and **URL** (`https://mcp.magnific.com`), then choose **OAuth** as the authentication method.
    5. Sign in to Magnific in the OAuth window that opens.
    6. **Start chatting.** ChatGPT will use Magnific whenever you ask for an image, a video, or an upscale.

    Developer-mode custom apps are available on Pro, Business, and Enterprise plans.
  </Tab>

  <Tab title="Claude Code">
    Add the Magnific MCP server with a single command:

    ```bash theme={null}
    claude mcp add --transport http magnific https://mcp.magnific.com
    ```

    The first call from inside Claude Code opens the OAuth flow in your browser. Once you approve it, the Magnific tools are available in the session.
  </Tab>

  <Tab title="Cursor">
    1. Open Cursor settings and go to the **MCP** tab.
    2. Click **Add new global MCP server**.
    3. Paste the configuration below:

    ```json theme={null}
    {
      "mcpServers": {
        "magnific": {
          "url": "https://mcp.magnific.com"
        }
      }
    }
    ```

    4. Reload the server entry. Cursor will open your browser for the Magnific sign-in.

    For more details, see [Cursor's MCP documentation](https://docs.cursor.com/context/model-context-protocol).
  </Tab>
</Tabs>

Other MCP-compatible clients (Windsurf, VS Code, OpenClaw, Hermes, Codex, and so on) follow the same idea. Add `https://mcp.magnific.com` as a streamable HTTP MCP server, name it `Magnific`, and finish the OAuth sign-in when it pops up.

## Available tools

These are the tools the MCP server exposes. The names are stable across clients, so you can reference them directly in your prompts.

### Account

* `account_balance`: current credit balance
* `project_report`: overview of project usage

### Creations

* `creations_search`: search your creations by query and filters
* `creations_get`: fetch a single creation by identifier
* `creations_show`: render creations inline in supported clients
* `creations_wait`: wait for an in-progress creation to finish
* `creation_status`: poll the status of an in-progress creation
* `creations_request_upload` / `creations_upload` / `creations_finalize_upload`: upload an asset
* `creations_move`: move a creation between folders

### Image generation and editing

* `images_generate`: generate images from text and optional references
* `images_generate_svg`: generate SVG output
* `images_to_svg`: convert raster to SVG
* `images_upscale`: Magnific upscaler
* `images_crop`: smart crop
* `images_resize`: resize
* `images_remove_background`: alpha cutout
* `images_models_list` / `images_models_show`: image model catalog

### Video generation

* `video_generate`: generate video
* `video_models_list` / `video_models_show`: video model catalog

### Audio

* `audio_tts`: text-to-speech
* `audio_voices_list` / `audio_voices_show`: voice catalog

### 3D

* `models3d_generate`: generate a 3D model

### Custom references

* `custom_references_create`: train a Soul character or style
* `custom_references_list`: list trained references

### Folders and Spaces

* `folders_list` / `folders_get` / `folders_create` / `folders_rename` / `folders_delete`
* `spaces_list`: list your Spaces
* `spaces_view`: inspect a single Space

### Discovery

* `tools_show`: surface the picker UI for available tools

The MCP server is the source of truth. The live `tools/list` response always reflects the latest set of tools available.

## Frequently asked questions

<AccordionGroup>
  <Accordion title="How does Magnific connect to AI agents?">
    Magnific runs a remote server that speaks the Model Context Protocol over streamable HTTP. Any MCP-compatible client can connect to `https://mcp.magnific.com` and call Magnific tools right from the chat.
  </Accordion>

  <Accordion title="Which agents are supported?">
    Magnific MCP works with Claude (Web and Claude Code), ChatGPT, Cursor, and any other client that supports the MCP streamable HTTP transport. That includes Windsurf, VS Code, OpenClaw, Hermes, and Codex.
  </Accordion>

  <Accordion title="Do I need an API key?">
    No. You sign in once with your Magnific account through OAuth and the client keeps the session. There's nothing to manage or rotate.
  </Accordion>

  <Accordion title="How does pricing work?">
    MCP tools draw from the same credit balance as the rest of Magnific. Each generation costs credits based on the model and resolution you pick. See [Pricing](/pricing) for the current rates.
  </Accordion>

  <Accordion title="Can I reference my previous generations?">
    Yes. The `creations_search`, `creations_get`, and `creations_show` tools let your agent browse your generation history and use any past image or video as the starting point for a new one.
  </Accordion>

  <Accordion title="What about the Magnific REST API?">
    The REST API documented on this site is still fully supported for direct, programmatic integrations using API keys. The MCP server is the path for AI assistants and chat-based agents.
  </Accordion>
</AccordionGroup>