QC videos, podcasts, and clips before upload — timestamped flags with agent-ready repair prompts.
Your agent can generate video. It can't tell you the audio cut out at 0:08, a frame froze at the stitch seam, or the captions are cropped. UploadCheck can — and hands back the exact timestamps to fix.
For anyone whose agent, pipeline, or editor produces media that ships to real people — AI-video pipelines, Shorts/Reels, podcasts, client deliverables.
An agent that writes code can run the tests. An agent that renders a video has no idea whether it's broken.
So the broken export ships: silence where narration should be, a frozen frame where two clips were stitched, lips out of sync, captions clipped by the platform UI, loudness well off target. Somebody notices in the comments.
UploadCheck is the check your agent calls before it publishes. Point it at the finished file. It returns PASS, WATCH, or BLOCK with per-flag timestamps and machine-actionable fixes — so the agent repairs only the flagged span and re-checks, instead of guessing or shipping.
npx -y @drantoniou/uploadcheck-mcp
claude mcp add --env UPLOADCHECK_API_BASE_URL=https://api.uploadcheck.app \
--env UPLOADCHECK_API_KEY=YOUR_KEY \
--transport stdio --scope user uploadcheck -- npx -y @drantoniou/uploadcheck-mcp
mcp.json){
"mcpServers": {
"uploadcheck": {
"command": "npx",
"args": ["-y", "@drantoniou/uploadcheck-mcp"],
"env": {
"UPLOADCHECK_API_BASE_URL": "https://api.uploadcheck.app",
"UPLOADCHECK_API_KEY": "YOUR_KEY"
}
}
}
}
codex mcp add uploadcheck \
--env UPLOADCHECK_API_BASE_URL=https://api.uploadcheck.app \
--env UPLOADCHECK_API_KEY=YOUR_KEY \
-- npx -y @drantoniou/uploadcheck-mcp
Get a free key at uploadcheck.app — 40+ gates, 200 checked minutes/month, no card.
Machine-readable install snippets for every client live in mcp-install.json.
Keep your key private. It's a secret bearer token tied to your plan minutes. Put it in your personal MCP config — never in a committed file, a shared prompt, or a report.
Once installed, just ask:
check ./final-cut.mp4 before I upload
The agent calls qc_run_local_file, polls qc_get_job, then reads qc_get_report and shows you the flags with timestamps.
40+ automated gates run on every check, including:
| Picture | frozen/looped frames, black frames, canvas fit & pillarboxing, temporal warping, flicker, interlace combing |
| Audio | dead air, clipping, loudness (LUFS) & true peak, loudness range, channel balance, mains hum, dropouts |
| Sync & text | lip/audio drift, caption safe-area, text contrast, cropped or overlapping text |
| Delivery | resolution/codec/frame-rate spec, Shorts/Reels format |
Paid plans add AI "oracle" gates — garbled speech, duplicated faces in AI crowds, narration match.
What it doesn't do: it doesn't tell you whether a clip was AI-generated. It assumes it might be, and checks whether it's clean enough to publish.
| Tool | What it does |
|---|---|
qc_estimate_cost | Estimate checked minutes and effective gates before running |
qc_run_local_file | Check a local export (reads the file, sends it to the hosted API) |
qc_run_video | Check a YouTube URL, signed URL, upload id, or base64 payload |
qc_get_job | Poll status, progress, verdict, metered minutes |
qc_get_report | The finished report: verdict + timestamped flags + evidence |
qc_get_events | Lifecycle events, so the agent can explain what ran |
qc_get_artifacts | List generated report artifacts |
qc_get_marker_csv | Editor marker CSV (drop flags straight into your NLE timeline) |
qc_create_upload_url | Signed upload URL for larger files |
Pass idempotency_key to qc_run_video when an agent might retry the same asset — the API returns the existing job instead of duplicating the run.
Use profile when you know the media type but don't want to hand-pick gates: auto, shorts, audio, thumbnail, generic_creator_video, and named delivery targets like youtube_shorts, podcast_delivery, broadcast_r128, ai_render_gate. Explicit checks always win.
export UPLOADCHECK_API_BASE_URL="https://api.uploadcheck.app"
export UPLOADCHECK_API_KEY="<workspace_api_key>"
This package is a thin MCP client over the hosted UploadCheck API — checks run server-side, not on your machine. Small and medium local files can be sent inline (base64) without durable storage; use signed upload URLs for large ones.
MIT — see LICENSE.
Source-derived launch command. Check the maintainer’s required arguments and credentials before running:
npx -y @drantoniou/uploadcheck-mcpMerge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.
{
"mcpServers": {
"app-uploadcheck-uploadcheck": {
"command": "npx",
"args": [
"-y",
"@drantoniou/uploadcheck-mcp"
]
}
}
}Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.
Claude Desktop setup reference@drantoniou/uploadcheck-mcpnpmapp.uploadcheck/uploadcheck works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.
~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.~/.cursor/mcp.jsonRestart Cursor for changes to take effect..vscode/mcp.jsonReload VS Code window for changes to take effect.~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect..mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.