-
Notifications
You must be signed in to change notification settings - Fork 0
MCP Server Specification
omerninyo edited this page Sep 29, 2026
·
1 revision
Language / שפה: English | עברית
This Request for Comments (RFC) specifies the architectural design, protocol interfaces, tool schemas, and integration strategy for turning RightSub into an official Model Context Protocol (MCP) Server.
The objective is to expose RightSub's core media discovery, FFmpeg stream extraction, SubRefine BiDi processing, and translation orchestrator natively to leading AI agents and desktop environments (e.g. Claude Desktop, Antigravity, Cursor, Windsurf, and ChatGPT).
-
Natural Language as the Ultimate Media UI:
- Rather than forcing non-technical users to master CLI flags or building a cumbersome cross-platform GUI, users interact via conversational dialog: "I downloaded a new film, inspect its subtitle tracks, fix reversed punctuation, and polish it against canon lore."
-
Local-First Media Processing (Zero Bandwidth Waste):
- Media containers are multi-gigabyte files (2 GB to 30 GB) that cannot and should not be uploaded over the network.
- The MCP server runs locally over
stdio, executing FFmpeg on local disks. Track extraction completes in ~0.2 seconds, and only tiny subtitle text payloads (~50 KB) are shared with the LLM.
-
Transparent Credential Management:
- Users interacting through their favorite desktop AI client leverage the client's existing LLM subscription without needing separate API key management.
-
Agentic Subtitle Quality Control:
- AI agents can invoke
inspect_media, analyze cues, detect franchise lore anomalies, and review diffs autonomously.
- AI agents can invoke
- Protocol Standard: Anthropic Model Context Protocol (MCP) Specification.
-
Transport Mechanism:
stdio(Standard Input / Standard Output) exchanging JSON-RPC 2.0 messages. -
Security Boundaries:
- Sandboxed subprocess under the host AI application.
- Strict filesystem scoping restricted to media directories.
- Read-only protection on video containers (
.mkv,.mp4).
-
Invocation Command:
rightsub mcp # or via python: python3 -m scripts.mcp_server
- Description: Probes a video file, subtitle file, or directory and returns structured media information.
-
Arguments:
-
path(string, required): Absolute filesystem path to media container or subtitle.
-
-
Returns:
- Embedded subtitle tracks (languages, codecs, track indices).
- External companion subtitle presence and encoding (UTF-8, CP1255).
- BiDi RLM compliance status.
-
Description: Extracts embedded subtitle stream to standalone
.srtfile. -
Arguments:
-
video_path(string, required): Path to video container. -
lang(string, default:"eng"): Preferred language code. -
track_index(integer, optional): Specific stream index.
-
- Description: Runs SubRefine engine to inject Plex/Infuse RLM marks, strip ads, and normalize character encodings.
-
Arguments:
-
subtitle_path(string, required): Path to subtitle file. -
in_place(boolean, default:false): Overwrite in-place with.baksafety.
-
- Description: Reviews and modernizes a batch of Hebrew cues against English master text using franchise domain rules.
-
Arguments:
-
english_cues(array of objects, required): Master English dialogue cues. -
hebrew_cues(array of objects, required): Hebrew translation cues. -
domain(string, optional): Domain pack ("sci_fi","legal","military", etc.).
-
Add to claude_desktop_config.json or Antigravity MCP settings:
{
"mcpServers": {
"rightsub": {
"command": "rightsub",
"args": ["mcp"],
"env": {
"PYTHONUNBUFFERED": "1"
}
}
}
}- Target Release: Version 1.5.0
-
Milestone:
v1.5.0 — Agentic Subtitle Protocol & MCP Server - Tracking Issue: #3
- Roadmap: Product Roadmap
RightSub Wiki — Subtitles Done Right. Powered by the SubRefine Algorithmic Engine & SubSwarm Multi-Agent AI.
- Home
- 📦 Global Installation Guide
- 🔰 Quickstart for Beginners
- BiDi & Plex Guide
- Pipeline Workflow
- On-Device STT & Sync (quicksubs)
- TMDb Metadata & Entity Resolution
- AI Assistants & Integration
- ⚖️ RightSub vs. Bazarr Comparison
- 🔄 Home Media & Download Integrations
- 🔮 Interactive CLI Specification
- 🔮 Setup & Health Wizard Specification
- 🔌 MCP Server Specification
- 💎 Semantic AI Polish & QC Specification
- 🐳 Docker Webhook Server Specification
- 🗺️ Product Roadmap
- 🚀 What's New & Release Notes
- Boston Legal Case Study
- CLI Reference
- דף הבית (Home HE)
- 📦 מדריך התקנה גלובלית והפצה
- 🔰 מדריך פשוט למתחילים
- מדריך BiDi ו-Plex
- תהליך עבודה מלא
- תמלול וסנכרון מקומי (quicksubs)
- אינטגרציית TMDb (עלילה ומגדר)
- חיבור לכלי בינה מלאכותית (AI)
- ⚖️ השוואה טכנית מול Bazarr
- 🔄 מדריך אינטגרציות ואוטומציה לשרתי מדיה
- 🔮 מפרט אשף פקודה אינטראקטיבי
- 🔮 מפרט אשף התקנה ואבחון
- 🔌 מפרט שרת MCP
- 💎 מפרט מנוע ליטוש סמנטי ו-QC
- 🐳 מפרט שרת Webhook וקונטיינר
- 🗺️ מפת דרכים ומעקב אבני-דרך
- 🚀 מה חדש ועדכוני גרסאות
- מקרה בוחן - בוסטון ליגל
- מדריך פקודות CLI