Transloadit

@transloadit list the last 10 assemblies from the past week and show me which ones failed or are still running

Found 10 assemblies from the past 7 days. 8 completed successfully, 1 is currently executing (video transcoding, 47% progress), and 1 failed with error ASSEMBLY_EXECUTING_FAILED during the /video/encode step. The failed assembly ID is 'd76c2e50ab…' created 3 days ago. All completed assemblies processed between 2-8 files each, with total processing times ranging from 12 seconds to 4 minutes.

This example uses the List Assemblies tool with date and state filters. Transloadit returns paginated results (default 50 per page); this prompt requests 10. The MCP cannot automatically retry failed assemblies—you must explicitly ask it to replay a specific assembly ID if you want to recover from a crash.

@transloadit start an assembly to resize profile-pic.jpg to 400x400 and convert it to WebP, then upload the result to my S3 bucket

Assembly 'a3f9d12e7c…' created and executing. Uploaded profile-pic.jpg (1.2 MiB). Processing steps: 1) /image/resize to 400×400 with smart crop, 2) /image/optimize converting to WebP with 85% quality, 3) /s3/store uploading to your configured bucket. Assembly URL: https://transloadit.com/assemblies/a3f9d12e7c… — you can poll this for completion or wait for the notify_url webhook if configured.

This example uses Create Assembly with inline Steps. The 20 MiB aggregate upload limit applies when attaching local files directly through the MCP—larger files require /http/import or /s3/import steps instead. The assembly runs asynchronously; the MCP returns immediately with the assembly ID, not the final output URLs.

@transloadit check assembly 'd76c2e50ab' that failed yesterday, tell me what went wrong, and replay it if the inputs are still available

Assembly 'd76c2e50ab…' failed at the /video/encode step with ASSEMBLY_EXECUTING_FAILED: 'FFmpeg encountered a corrupt frame at 00:03:12'. Original input video.mp4 (450 MiB) is still recoverable. Replayed assembly as new ID 'b8e4a19f2d…' using the same Template. The new assembly is now queued—Transloadit will reprocess from scratch. Monitor b8e4a19f2d… for completion; if it fails again, the input file may be inherently corrupt.

This example chains Get Assembly Status with Replay Assembly. Replay only works if Transloadit still has the original uploads cached (typically 24 hours). The replayed assembly gets a new ID and counts as a separate billable job. If the underlying file is corrupt, replay will fail identically—Transloadit cannot fix broken source media.

When Transloadit beats building your own media pipeline

A 6-person SaaS team needs to accept user-uploaded videos, transcode them to three resolutions, and store thumbnails in S3. Transloadit wins here because the Create Assembly tool lets you define the entire pipeline as a single API call—upload, transcode, thumbnail, store—and the Get Assembly Status tool gives you real-time progress without polling infrastructure. The 20 MiB direct-upload limit means files over that size need a separate upload step, but most profile photos and short clips fit. If your app processes fewer than 500 uploads a day, this MCP keeps your team out of FFmpeg hell. Use Switchy to wire the Assembly webhook into your Slack or Linear board so the whole team sees when a batch fails.

Replaying failed uploads when customers complain

A 3-person support team at a content platform gets tickets about failed video uploads. The List Assemblies tool filters by date and keyword, so an agent can pull up a customer's last five attempts in one query. If the Assembly crashed mid-transcode, the Replay Assembly tool retries it without asking the customer to re-upload—Transloadit keeps the original file for 24 hours. This scenario works when your support volume is under 50 tickets a day and most failures are transient encoding errors, not corrupt files. If you're debugging hundreds of uploads daily, you need a dedicated observability stack. Use Switchy to let support agents run these lookups without touching the Transloadit dashboard or bothering engineering.

When you need to reprocess 10,000 assets overnight

A 4-person marketing team is rebranding and needs to regenerate watermarked thumbnails for 10,000 archived videos. The List Assemblies tool pages through old jobs, and Replay Assembly Notification resends results to a new webhook once the batch finishes. Transloadit's rate limit is 3 status checks per second per Assembly, so you can't hammer the API, but the MCP's Cancel Assembly tool lets you stop runaway jobs before they burn your processing credits. This scenario breaks down if your archive is over 50,000 assets—you'll hit pagination fatigue and need a bulk-export script. Use Switchy to coordinate the migration across marketing, eng, and ops so everyone sees which batches are stuck.