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

# Overview

> Connect MCP-compatible AI agents to VMEG media translation

export const mcpUrl = "https://www.vmeg.ai/api/mcp";

export const apiKeyUrl = "https://www.vmeg.ai/open-api-setting";

export const productUrls = JSON.parse('{"home":"https://www.vmeg.ai/home","apiConfiguration":"https://www.vmeg.ai/open-api-setting","signIn":"https://www.vmeg.ai/signIn?redirect=%2Fhome","signInToApiSettings":"https://www.vmeg.ai/signIn?redirect=%2Fopen-api-setting","signUp":"https://www.vmeg.ai/signup","account":"https://www.vmeg.ai/account","helpCenter":"https://www.vmeg.ai/help-center/","pricing":"https://www.vmeg.ai/pricing","myTasks":"https://www.vmeg.ai/my-tasks","myAssets":"https://www.vmeg.ai/my-assets","videoTranslation":"https://www.vmeg.ai/translation","audioTranslation":"https://www.vmeg.ai/audio-translation","cloneVoice":"https://www.vmeg.ai/clone-voice","bookDemo":"https://www.vmeg.ai/book-a-demo/","contactUs":"https://www.vmeg.ai/contact-us/","apiDocs":"https://docs.vmeg.ai"}');

Connect any **MCP-compatible** AI agent to VMEG by **manually** adding the remote MCP endpoint and your <a href={apiKeyUrl}>API Key</a>. Your agent can list materials, create translation tasks, edit drafts, and poll job status — using the same account credits as the website.

<Info>
  **Marketplace / app-store connectors are not supported yet.** Use [manual setup](/guides/mcp/setup) with an <a href={apiKeyUrl}>API Key</a> for now.
</Info>

## Remote MCP endpoint

Point your MCP client at:

```text theme={null}
{mcpUrl}
```

The server uses **Streamable HTTP** (Spring AI MCP Server, sync tools).

## Authentication

**Recommended today:** if your client supports a Bearer token, send your <a href={apiKeyUrl}>API Key</a> on every MCP request (same key as the [Open API](/guides/authentication)):

```http theme={null}
Authorization: Bearer YOUR_API_KEY
```

Create and rotate keys on <a href={apiKeyUrl}>API Configuration</a>.

| Mode                 | Status              | Notes                                                                                                     |
| -------------------- | ------------------- | --------------------------------------------------------------------------------------------------------- |
| **API Key (Bearer)** | Supported           | Manual MCP config in Cursor or any compatible client. Key from <a href={apiKeyUrl}>API Configuration</a>. |
| **OAuth 2.1 (PKCE)** | Available on server | For future connector flows; not needed for manual setup.                                                  |

<Info>
  You do **not** pass `projId` in MCP tool arguments. The authenticated project comes from your <a href={apiKeyUrl}>API Key</a> (or OAuth token when using connectors later).
</Info>

## What you can do

Once connected, your agent can call tools for:

* **Materials** — list, upload (`POST /api/mcp/material/upload`), delete
* **Tasks** — create video/audio/subtitle translation, list and poll status, delete
* **Voices** — list system and cloned voices for manual voice selection
* **Editor workflows** — query scripts/subtitles/ASR/speakers/tracks, save draft edits, re-translate, trigger TTS, compose final render

See the full [MCP tools reference](/guides/mcp/tools) for every tool name and description.

## Typical flow

<Steps>
  <Step title="Connect manually">
    Add `{mcpUrl}` to your MCP client and set `Authorization: Bearer` plus your <a href={apiKeyUrl}>API Key</a>. See [Setup](/guides/mcp/setup).
  </Step>

  <Step title="Provide media">
    Upload a file (`POST /api/mcp/material/upload`), pick from `vmeg_list_materials`, or pass a YouTube URL.
  </Step>

  <Step title="Create a task">
    Call `vmeg_create_video_translation_task` or `vmeg_create_subtitle_translation_task` with `targetLanguage` and a media source.
  </Step>

  <Step title="Poll and edit">
    Use `vmeg_list_tasks` and editor tools to query content, save draft changes, re-dub, or compose the final output.
  </Step>
</Steps>

Tasks are **asynchronous**. Creating a job returns a `taskId` and `editorUrl`; poll until the task finishes.

## Remote MCP vs Open API

|          | Remote MCP                                       | [Open API](/guides/introduction)     |
| -------- | ------------------------------------------------ | ------------------------------------ |
| Setup    | Manual MCP URL + <a href={apiKeyUrl}>API Key</a> | <a href={apiKeyUrl}>API Key</a> only |
| Host     | `www.vmeg.ai` (`/api/mcp`)                       | `api.vmeg.ai` (`/openapi/v1/**`)     |
| Best for | AI agents (conversational workflows)             | Your backend (REST integration)      |
| Billing  | Account credits (same plan)                      | Account credits (same plan)          |

Both access the same VMEG project, materials, and tasks.

## FAQ

**Do I need a separate MCP API Key?**\
No. Reuse the same <a href={apiKeyUrl}>API Key</a> as the Open API — send it as the Bearer token.

**Can I install VMEG from the Cursor marketplace or ChatGPT app store?**\
Not yet. Use [manual setup](/guides/mcp/setup) until marketplace connectors are published.

**Does MCP cost extra?**\
No. Tool calls consume the same credits as tasks created on <a href={productUrls.home}>vmeg.ai</a>. See [Pricing](/guides/pricing).

**Can I use MCP and the Open API together?**\
Yes. They operate on the same project scope; choose MCP for agents and REST for programmatic backends.

**Why does the agent ask me to confirm before delete or compose?**\
Delete tools (`vmeg_delete_material`, `vmeg_delete_task`) and destructive editor actions require explicit user confirmation by design.