File hosting · curl API · agent-readable
Upload a file from any shell, script, CI job, or AI agent and get a public URL back. No account, no browser step, no ad page in front of the download.
curl -sf -F "file=@report.pdf" "https://d.iboost.ua/api/anon/upload?format=text" https://d.iboost.ua/f/AbCd1234/report.pdf
That is the whole flow. The response is the link itself — one line, ready to paste into a message or hand back to whoever asked for the file.
Upload in the browserCoding agents waste a surprising amount of effort on file hosts: probing endpoints, trying scp, scraping upload forms, hunting for a delete API that does not exist. So this service publishes its own machine-readable spec at /llms.txt — every endpoint, every limit, the response shape, and an explicit list of dead ends not to try.
An agent fetches one URL and knows how to work. The bare domain returns the same document to command-line clients, so even a blind curl https://d.iboost.ua lands on the instructions instead of a 404.
Anonymous uploads work immediately — nothing to register, nothing to configure in a CI job.
POST multipart or PUT a raw body. Works from curl, Python, Go, a Makefile, a cron job.
The URL serves the file itself. No interstitial, no countdown, no ads, no tracking of who clicked.
?format=text returns the bare URL for shell pipelines; the default returns JSON with size, mime, expiry and a management link.
| anonymous | with a free key | |
|---|---|---|
| max file size | 200 MB | 1 GB |
| lifetime | 7 days | no expiry |
| storage | shared pool | 5 GB per account |
| rate | 100 uploads per hour per IP | |
A key is created in the panel and passed as Authorization: Bearer dfh_…. The same key works with ShareX and CleanShot, so screenshots land here too.
curl -sf -H "Authorization: Bearer $DFILE_TOKEN" \
-T dump.sql "https://d.iboost.ua/api/dump.sql"
Uploaded files are never indexed — search engines and crawlers are refused on every file path, and nothing is listed publicly. A link is private until you share it, and after that it is as public as the people you gave it to. Files are virus-scanned on arrival. There are no ads and no third-party analytics on download pages.
It is not a backup service and not a CDN: anonymous files expire on schedule, and there is no replication behind them. Keep your own copy of anything that matters.
No. The anonymous endpoint works with no registration at all — that is the intended path for scripts and agents. An account only raises the file size to 1 GB and removes the 7-day expiry.
Every upload returns a manage link in its JSON response. Opening it gives you rights over that file — rename, extend, or delete it. Save it if the file may need to be removed.
Yes — that is the main use case. Both endpoints are plain HTTP with no CSRF, no cookies and no interactive step, over IPv4 and IPv6.
It is a plain-text spec written for machines rather than people. AI coding agents that are told to "upload this somewhere" otherwise explore blindly; here they read /llms.txt once and use the correct endpoint on the first try.
On a single server in Europe, operated by iBoost. No third-party object storage is involved, and no file leaves that machine.
Use the report form — paste the link and what is wrong with it. Malware, phishing and copyright-infringing uploads are removed on report, usually the same day. You can also write to hello@iboost.ai.