Give your agent a clear editing plan.

ربط الموقع بالوكلاء

انسخ رابط MCP أدناه إلى عميل يدعم Streamable HTTP. الأدوات تعمل دون مفتاح API. الموصل يجهّز خطة القصّ ويقرأ نص الترجمة الذي ترسله؛ اختيار الفيديو والمراجعة والتصدير تتم داخل المحرر على جهازك.

الحسابات الاجتماعية غير مربوطة، والنشر المباشر غير مفعّل. توفر الرابط لا يعني تثبيت التطبيق تلقائيًا أو اعتماده في متجر ChatGPT.

Connection

https://maher-cut-studio.maherr20.chatgpt.site/mcp

Transport: Streamable HTTP · Authentication: none · Version: 1.1.0

The endpoint accepts MCP JSON-RPC POST requests with Content-Type: application/json and Accept: application/json, text/event-stream. It is stateless, with no persistent session and no standalone SSE stream. Use an MCP client to initialize, discover tools and call them.

{
  "mcpServers": {
    "laqta": {
      "url": "https://maher-cut-studio.maherr20.chatgpt.site/mcp"
    }
  }
}

Client configuration formats vary. Add this as a custom remote MCP server where your client supports it. No installation or marketplace approval is implied.

Available tools

• get_capabilities — read implemented features, privacy, limits and editor links.
• prepare_edit_plan — validate retained intervals against the known video duration; return JSON, output duration and an optional link that preloads timestamps.
• parse_subtitles — parse supplied SRT/VTT into timestamped cues. This does not transcribe audio.

The resource laqta://capabilities exposes the same capability data. All remote tools process their supplied arguments without storing a project or changing a phone session.

Example: keep three parts

When the user has explicitly chosen these ranges in a 50-minute source, call prepare_edit_plan with:

{
  "duration": 3000,
  "clips": [
    {"start": 120, "end": 150, "label": "Opening"},
    {"start": 600, "end": 900, "label": "Main explanation"},
    {"start": 1500, "end": 1890, "label": "Demonstration"}
  ]
}

The result is a 720-second edit plan. These are example timestamps, not an analysis of any real video. Only use timestamps grounded in the user's instructions or a supplied timed transcript.

1. Copy plan_json from the result.
2. Open the editor and select the original video.
3. Expand “تطبيق خطة المونتاج”, paste the JSON, then apply.
4. Review the cuts, export, and save the resulting file.

The optional preview_cuts_url preloads only timestamps, with no transcript, title, tags or filename. It still requires choosing a video and manually applying the plan. Source-duration checking cannot prove that two files with the same duration are identical; include the exact optional source identity from the editor for stricter matching.

Limits and privacy

Up to 150 clips, 150,000 subtitle characters and 3,000 parsed cues per request. Invalid subtitle blocks are skipped; text is limited to 500 characters per cue. Request bodies are limited to 300,000 bytes.

No video upload, remote rendering, speech recognition, automatic footage analysis, social OAuth or social publication is implemented. Browser export is experimental and takes roughly the output duration; long phone exports are not guaranteed.

Local footage stays in the browser. MCP arguments go to the hosting service for processing; the application does not persist them. Hosting and client logging policies may apply. No paid AI API is invoked.