Quick start
Every product uses your API key. Send it as Authorization: Bearer YOUR_KEY (API and crawl) or as the proxy password.
| Product | Address |
|---|---|
| AI and storage API | https://api.depined.org/v1 |
| Crawl API | https://hub.depined.org/fetch |
| Proxy (HTTP) | proxy.depined.ai:18080 |
| Proxy (HTTPS) | proxy.depined.ai:18443 |
| Proxy certificate | depined-ca.zip (or the .crt file) (only for browser-extension exits, see below) |
curl "https://hub.depined.org/fetch?url=https://api.ipify.org" \
-H "Authorization: Bearer {{KEY}}"Routing: choose the devices
By default each request goes to the least busy device and the IP rotates. Use these options to choose. They work the same on the crawl API (query parameters) and the proxy (in the username).
| Option | Crawl API | Proxy username | Meaning |
|---|---|---|---|
| Device type | type=desktoptype=desktop,mobile | -type-desktop-type-desktop_mobile | desktop (desktop app), extension (Chrome extension), mobile (phone app), or any |
| Country | country=UScountry=US,GB,DE | -country-us-country-us_gb_de | Exit country (ISO code). Several = any of them |
| Sticky IP | session=abc123 | -session-abc123 | Same device (IP) for every request with this id, kept for 30 minutes of activity. Letters and digits, up to 32 |
Every response says which device served it: x-depined-node-type, x-depined-country, x-depined-node. If no device matches, you get 503 with the reason (for example no mobile nodes online in US); loosen the filter or retry.
Proxy
A standard HTTP / HTTPS proxy for browsers, scrapers and apps. Username: client plus routing options; password: your API key.
curl -x "http://client-country-us:{{KEY}}@proxy.depined.ai:18080" https://api.ipify.orgimport requests
proxy = "http://client-type-desktop-country-us-session-a1:{{KEY}}@proxy.depined.ai:18080"
r = requests.get("https://api.ipify.org", proxies={"http": proxy, "https": proxy},
verify="depined-ca.crt") # only needed for extension exits
print(r.text)import { ProxyAgent, fetch } from "undici";
const dispatcher = new ProxyAgent("http://client-country-de:{{KEY}}@proxy.depined.ai:18080");
const r = await fetch("https://api.ipify.org", { dispatcher });
console.log(await r.text());The certificate: when you need it
- Desktop app and phone exits (
-type-desktop,-type-mobile): a real tunnel. Your TLS goes straight to the site; no certificate needed, any HTTPS client works. - Chrome-extension exits (
-type-extension, or "any" when no tunnel device is free): the proxy opens HTTPS for the browser to fetch, so your client must trust the Depined certificate.
| Where | How to trust it |
|---|---|
| curl | --cacert depined-ca.crt |
| Python requests | verify="depined-ca.crt" |
| Node.js | NODE_EXTRA_CA_CERTS=depined-ca.crt node app.js |
| macOS | sudo security add-trusted-cert -d -r trustRoot -k /Library/Keychains/System.keychain depined-ca.crt |
| Windows | certutil -addstore -f Root depined-ca.crt (as administrator) |
| Linux (Debian/Ubuntu) | sudo cp depined-ca.crt /usr/local/share/ca-certificates/ && sudo update-ca-certificates |
Only web ports (80, 443, 8080, 8443) are carried, and private / local addresses are always refused.
Crawl API
Fetch any public URL through a member's device and get the site's response as is (status, headers, body). GET and POST are supported; your request headers and body are passed on.
curl -i "https://hub.depined.org/fetch?url=https://example.com&type=desktop&country=US" \
-H "Authorization: Bearer {{KEY}}"| Parameter | |
|---|---|
url | Page to fetch (required, URL-encoded) |
type, country, session | Routing, see above |
Request bodies (for POST) up to 700 KB. If a device fails before answering, the request is retried on up to 3 devices. No certificate is ever needed for the crawl API.
AI models (OpenAI-compatible)
Point any OpenAI client at https://api.depined.org/v1 with your key. List what is available with GET /v1/models.
from openai import OpenAI
client = OpenAI(base_url="https://api.depined.org/v1", api_key="{{KEY}}")
r = client.chat.completions.create(model="MODEL_ID", messages=[{"role": "user", "content": "Hello!"}])
print(r.choices[0].message.content)| Endpoint | What |
|---|---|
POST /v1/chat/completions | Chat / text models |
POST /v1/embeddings | Text embeddings |
POST /v1/images/generations | Images: prompt, size, n, steps, seed |
POST /v1/videos/generations | Video (split across several members' GPUs): prompt, size, frames. Takes minutes; use a long timeout |
GET /v1/jobs/{id} | Status and progress of a running job |
POST /v1/tasks | Container workloads: {"type": "...", "input": {...}}; list them with GET /v1/tasks/types |
Storage
Files are encrypted and split into redundant pieces held by members' devices; any part of the network can go offline and the file still comes back.
curl -X POST "https://api.depined.org/v1/storage?name=report.pdf" \
-H "Authorization: Bearer {{KEY}}" --data-binary @report.pdf
curl "https://api.depined.org/v1/storage" -H "Authorization: Bearer {{KEY}}"
curl -o report.pdf "https://api.depined.org/v1/storage/FILE_ID" -H "Authorization: Bearer {{KEY}}"
curl -X DELETE "https://api.depined.org/v1/storage/FILE_ID" -H "Authorization: Bearer {{KEY}}"Up to 100 MB in one request. Bigger files (up to 1 TB) use chunked upload: POST /v1/storage/uploads?name=&size=, then PUT /v1/storage/uploads/{id}/parts/{n} (64 MB parts), then POST /v1/storage/uploads/{id}/complete. The helper script depined_storage.py does it for you.
Errors and limits
| Status | Meaning |
|---|---|
401 | Missing or wrong API key |
403 | Key disabled, or a private / blocked target |
407 | Proxy: username / password missing |
502 | Devices failed to reach the site (we tried several) |
503 | No device matches your routing right now |
504 | The device took too long |
Questions or higher limits: hello@depined.org.