VideoDB Skills

Why add this Skill

This skill gives your agent one consistent interface to:

• See: Realtime desktop screen, mic and system audio, RTSP streams, ingest files, URLs, YouTube.

• Understand: Run analyzers over speech, scenes, objects, OCR and brands; index the results; then search moments, ask questions, filter and count

• Act: Stream results, trigger alerts on live feeds, edit timelines, generate subtitles and overlays, export clips.

What it does

VideoDB Skills lets your AI coding agent run end to end, server-side video workflows in real time and batch:

• Capture desktop screen, mic, and system audio for real time processing.
• Upload and process RTSP streams, videos from YouTube, URLs, and local files.
• Create realtime context of visual and spoken information.
• Run analyzers over speech, scenes, objects, OCR and brands, then index what they produce.
• Search moments in natural language, ask questions and get answers grounded in the video, filter on exact field values, and aggregate counts server-side.
• Generate transcripts, subtitles, and AI media.
• Edit clips, overlays, and exports server side.

Return playable HLS links for anything you build.

Get Started

Get started in two quick steps. Open your AI coding agent (Requires Python 3.9+ and VideoDB Python SDK 0.5.0 or newer) and follow along.

Step 1: Install the skill

npx skills add video-db/skills

Or install with Claude Code plugin:

/plugin marketplace add video-db/skills
/plugin install videodb@videodb-skills

The Claude Code plugin includes the Python skill and the hosted MCP server at https://mcp.videodb.io/mcp. After installation, run /mcp, select VideoDB, and complete browser authorization with your VideoDB account. Then ask Claude to "list my VideoDB collections" to verify the connection. The hosted MCP tools do not require a local Python installation or a manually configured API key.

The Python skill uses separate SDK authentication; follow Step 2 below for SDK workflows. Installing with npx skills add installs the skill only, without the MCP connection. See the MCP setup guide and privacy policy.

Step 2: Setup

/videodb setup

For the Claude Code plugin, use the fully qualified command /videodb:videodb setup.

The agent will guide setup for your VideoDB API key ($20 free credits, no credit card required), install the SDK, and verify the connection.

For Cursor, Copilot, and other agents, ask your agent to "setup videodb"

Set your API key using either method:

# Recommended: Export in terminal
export VIDEO_DB_API_KEY=sk-xxx

# Or add to your project's .env file
VIDEO_DB_API_KEY=sk-xxx

---

Give your agent instructions

Ask your agent to run instructions like these. The skill loads automatically.

• "Upload https://www.youtube.com/watch?v=MnrJzXM7a6o and give me a sharable stream link"
• "Take clips from 10s-30s and 45s-60s and compile them"
• "Generate a background music, and add to this Clip"
• "Add subtitles to original video with white text on black background"
• "Find every scene showing 'phone close-up' or 'product on screen'"
• "Which objects appear most often in this video? Break it down by shot count"
• "Ask this video what the presenter said about pricing, and show me the clips"
• "Capture my screen for the next two minutes and write a report of what i'm doing along with any insights or suggestions"*
• "Here is the rtsp link for my IP Camera , monitor and log the alert to text file along with timestamp whenever a person enters into the room"

---

Capability

VideoDB is the server side video stack for agents and apps. Run reliable, scalable, cost efficient workflows across realtime streams and batch video, with built in AI understanding, without wiring up ffmpeg glue. Keep your client and agent stack light: send video in, get back structured context, searchable moments, and playable streams.

When to use VideoDB

• Your app needs video workflows, but you do not want ffmpeg running everywhere
• You want realtime perception from RTSP feeds or desktop capture
• You need search by what was said or shown, then turn results into clips
• You want server side editing, reframing, subtitles, dubbing, and streaming links

Capability	What it unlocks
Capture	Capture desktop screen, mic, and system audio for realtime processing
Upload	Ingest video from YouTube, URLs, or local files
Context	Generate realtime structured context for any RTSP feed or desktop stream
Understand	Run analyzers (speech, scenes, objects, OCR, brands, activity) into reusable artifacts
Search	Find exact moments by speech, scenes, or metadata, return playable evidence
Ask & Analyze	Get grounded answers with evidence, filter on exact fields, aggregate counts and facets
Transcripts	Generate clean, timestamped transcripts from any video
Subtitles	Auto generate subtitles, then style and burn in or export
Edit	Trim, merge, clip, overlay text, images, audio, plus dubbing and translation
AI Generate	Create images, video, music, sound effects, and voiceovers from text
Sandbox Compute	Run open-weight models (Gemma, Qwen, Whisper, OmniVoice, FLUX, RT-DETR) on managed GPU sandboxes
Transcode / Reframe	Change resolution, quality, aspect ratio, and social crops, all on the server
Stream	Get instant playable HLS links (built in CDN) for anything you ingest or generate.

The idea in one line

See → Understand → Act, as an API, for video and audio.

Supported Platforms: macOS, Linux, Windows (PowerShell)

---

Community & Support

• Documentation: docs.videodb.io
• Discord: Join our community