YouTube Shorts Agent
Agent-first YouTube Shorts uploader with dry-run validation, OAuth checks and synthetic-media metadata.
README
<!-- delx header v2 --> <h1 align="center">YouTube Shorts Agent</h1>
<div align="center"> <img src="assets/banner.png" alt="YouTube Shorts Agent" width="85%" /> </div>
<h3 align="center"> Agent-first YouTube Shorts uploader for the YouTube Data API.<br>Dry-run safe, OAuth readiness checks, MCP-ready for Codex / Claude / Cursor / Hermes. </h3>
<p align="center"> <a href="https://www.npmjs.com/package/youtube-shorts-agent"><img src="https://img.shields.io/npm/v/youtube-shorts-agent?style=for-the-badge&labelColor=0F172A&color=10B981&logo=npm&logoColor=white" alt="npm version" /></a> <a href="https://www.npmjs.com/package/youtube-shorts-agent"><img src="https://img.shields.io/npm/dm/youtube-shorts-agent?style=for-the-badge&labelColor=0F172A&color=0EA5A3&logo=npm&logoColor=white" alt="npm downloads" /></a> <a href="LICENSE"><img src="https://img.shields.io/badge/LICENSE-MIT-22C55E?style=for-the-badge&labelColor=0F172A" alt="License MIT" /></a> <a href="https://modelcontextprotocol.io"><img src="https://img.shields.io/badge/BUILT_FOR-MCP-7C3AED?style=for-the-badge&labelColor=0F172A" alt="Built for MCP" /></a> </p>
<p align="center"> <a href="https://github.com/davidmosiah/youtube-shorts-agent/stargazers"><img src="https://img.shields.io/github/stars/davidmosiah/youtube-shorts-agent?style=for-the-badge&labelColor=0F172A&color=FBBF24&logo=github" alt="GitHub stars" /></a> <a href="https://github.com/davidmosiah/youtube-shorts-agent/actions/workflows/ci.yml"><img src="https://github.com/davidmosiah/youtube-shorts-agent/actions/workflows/ci.yml/badge.svg" alt="CI status" /></a> <a href="https://github.com/davidmosiah"><img src="https://img.shields.io/badge/PART_OF-Delx_Agent_Stack-0EA5A3?style=for-the-badge&labelColor=0F172A" alt="Part of the Delx agent stack" /></a> <a href="https://github.com/davidmosiah/youtube-shorts-agent"><img src="https://img.shields.io/badge/CATEGORY-Reach-FF0000?style=for-the-badge&labelColor=0F172A" alt="Category" /></a> </p>
⭐ If this agent-first tool helps your workflow, please star the repo. Stars make this tooling easier for other builders to discover and help Delx keep shipping open infrastructure.<br> 🧱 Part of the Delx agent stack — 15 open-source MCP servers across body, reach and coordination.
<!-- /delx header v2 -->
Agent-first YouTube Shorts uploader for the YouTube Data API. It is designed for Codex, Claude, Cursor, Hermes, OpenClaw and any MCP client that needs a predictable upload workflow with dry-run safety, OAuth readiness checks and structured output.
Use it when an agent needs to prepare, validate or upload Shorts through the official API without touching YouTube Studio UI.
What Agents Get
youtube_agent_manifestfor install/runtime guidanceyoutube_connection_statusbefore upload attemptsyoutube_privacy_auditfor local token and media boundariesyoutube_oauth_authorize_urlwith local PKCE session storageyoutube_upload_shortwithcontainsSyntheticMediasupportyoutube_list_recent_videosfor lightweight post-upload checks
Install
npm install -g youtube-shorts-agent
Or run directly:
npm exec --yes --package=youtube-shorts-agent -- youtube-shorts-agent doctor
CLI
youtube-shorts-agent manifest --client codex
youtube-shorts-agent doctor
youtube-shorts-agent privacy-audit
youtube-shorts-agent auth-url --redirect-uri http://localhost:8787/callback
youtube-shorts-agent upload-short --video ./short.mp4 --title "Launch title" --caption-file copy.txt
youtube-shorts-agent list-recent --max-results 10
Dry-run is enabled by default. Set YOUTUBE_DRY_RUN=false only when doctor reports a complete OAuth setup and you intend to call the live API.
First Upload (dry-run walkthrough)
This is the full first-run path. Every step here is dry-run safe — no credentials are required and nothing is sent to YouTube. The output below is captured verbatim from a real run.
1. Check readiness. With no OAuth configured, doctor confirms you are in dry-run mode:
youtube-shorts-agent doctor
{
"ok": true,
"dry_run": true,
"configured": {
"client_credentials": "missing",
"access_token": "missing",
"refresh_token": "missing"
},
"missing_count": 3,
"ready_for_live_upload": false,
"next_steps": [
"Current mode is dry-run. Validate metadata and agent flow before live uploads."
]
}
2. Prepare a Short and its caption.
printf 'Launching the agent-first Shorts uploader.\n#shorts #ai #agents' > copy.txt
# short.mp4 is your vertical 9:16 clip
3. Run the upload in dry-run. No network call is made; the tool returns the exact job and result it would publish, so an agent can validate metadata before going live:
youtube-shorts-agent upload-short \
--video ./short.mp4 \
--title "Agent-first Shorts upload" \
--caption-file copy.txt \
--tags ai,agents \
--duration 24
{
"ok": true,
"dry_run": true,
"job": {
"id": "youtube_1780082215631",
"platform": "youtube",
"status": "queued",
"createdAt": "2026-05-29T19:16:55.631Z",
"caption": "Launching the agent-first Shorts uploader.\n#shorts #ai #agents",
"targetUrl": "",
"mediaPaths": ["./short.mp4"],
"metadata": {
"title": "Agent-first Shorts upload",
"youtube_title": "Agent-first Shorts upload",
"youtube_tags": ["ai", "agents"],
"youtube_contains_synthetic_media": true,
"video_duration_sec": 24,
"video_aspect_ratio": "9:16"
}
},
"result": {
"provider": "youtube_official",
"platformPostId": "dryrun_1780082215631",
"releaseUrl": "https://www.youtube.com",
"raw": { "dryRun": true, "jobId": "youtube_1780082215631" }
}
}
platformPostId is prefixed with dryrun_ and releaseUrl is the YouTube root — both signal that nothing was uploaded. The id, createdAt and dryrun_* values vary per run.
4. Confirm the channel listing path (returns an empty list in dry-run):
youtube-shorts-agent list-recent --max-results 5
{ "items": [] }
Going live. Configure OAuth (youtube-shorts-agent auth-url --redirect-uri http://localhost:8787/callback, then exchange the callback code for tokens), set YOUTUBE_CLIENT_ID / YOUTUBE_CLIENT_SECRET / YOUTUBE_ACCESS_TOKEN / YOUTUBE_REFRESH_TOKEN in .env, re-run doctor until ready_for_live_upload is true, then set YOUTUBE_DRY_RUN=false and re-run the same upload-short command.
MCP
youtube-shorts-mcp
HTTP transport:
YOUTUBE_MCP_TRANSPORT=http youtube-shorts-mcp
Hermes-style config:
mcp_servers:
youtube_shorts:
command: npx
args: ["-y", "youtube-shorts-agent"]
sampling:
enabled: false
Recommended first calls:
youtube_connection_statusyoutube_privacy_audityoutube_upload_short
Agent Surfaces
| Tool | Purpose |
|---|---|
youtube_agent_manifest |
Install/runtime guidance for Codex, Claude, Cursor, Hermes and OpenClaw |
youtube_connection_status |
OAuth and dry-run readiness without token values |
youtube_privacy_audit |
Upload scope, synthetic media and local file boundaries |
youtube_oauth_authorize_url |
PKCE authorization URL with local session storage |
youtube_upload_short |
Dry-run or live Shorts upload |
youtube_list_recent_videos |
Lightweight channel verification |
Copy-Paste Agent Prompt
Use youtube-shorts-agent. First call youtube_connection_status and youtube_privacy_audit.
If uploading AI-generated media, keep containsSyntheticMedia=true. Never print token values.
Configuration
Copy .env.example to .env. Keep .env and .agent-data/ out of Git.
The upload tool sets containsSyntheticMedia=true by default for AI-generated or AI-edited videos. Override only when that is not true for the asset.
Safety Model
- OAuth tokens are never returned by CLI or MCP tools.
- PKCE verifier is stored locally in
.agent-data/. - Live upload requires
YOUTUBE_DRY_RUN=false. - The package uses the official YouTube Data API and does not automate Studio UI.
Development
npm install
npm test
npm run check
📧 Contact & Support
- 📨 support@delx.ai — general questions, integration help, partnerships
- 🐛 Bug reports / feature requests — GitHub Issues
- 🐦 Updates — @delx369 on X
- 🌐 Site — wellness.delx.ai
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.