What 200 working Telegram proxies look like
Our checker tests public MTProto proxies around the clock and ranks the ones that keep working. On 9 October 2026 we took the top 200 and looked at where they run and how they are built. Half of them sat with one German hosting company, and nearly a third ran inside Iran.
- 57% of working proxies ran in Germany, 50.5% with a single provider, Hetzner.
- 29.5% ran on servers inside Iran, the country where Telegram has been blocked since 2018. One Iranian network alone hosted 24%.
- Only 17.5% used port 443. 76% used the alternative HTTPS ports 8443, 8880, 2096 and 2053.
- 5% used FakeTLS, the hardest type to detect; none of the 59 Iranian-hosted ones did.
- 39 used a domain instead of an IP address, and 27 of those domains were named for Iranian users.
Where they are hosted
The 200 proxies ran on 25 networks in 11 countries. Germany is a common choice for cheap servers with fast international links, and 50.5% of the whole list sat with one provider. That concentration is a weakness: blocking that one provider's address ranges would cut off half of the proxies in our list at once.
| Hosting country | Proxies | Share |
|---|---|---|
| Germany | 114 | 57% |
| Iran | 59 | 29.5% |
| France | 6 | 3% |
| Hong Kong | 6 | 3% |
| United States | 5 | 2.5% |
| United Kingdom | 3 | 1.5% |
| Netherlands | 3 | 1.5% |
| Other | 4 | 2% |
The second cluster is inside Iran itself: 59 proxies, 48 of them on one Iranian network. These work as relays: users reach them over Iran's domestic network, and the server passes the traffic on to Telegram abroad. It is a practical design with a privacy cost, because the user's IP address lands on a machine under Iranian jurisdiction. Is a free proxy safe?
Which ports they use
Only 17.5% listened on 443, the standard HTTPS port. The favourite was 8443 (53.5%), followed by 8880, 2096 and 2053. These are ports that firewalls usually leave open for web traffic, so a proxy on them does not stand out the way an odd port like 54321 would.
| Port | Proxies | Share |
|---|---|---|
| 8443 | 107 | 53.5% |
| 443 | 35 | 17.5% |
| 8880 | 20 | 10% |
| 2096 | 14 | 7% |
| 2053 | 11 | 5.5% |
| 444 | 4 | 2% |
| Other | 9 | 4.5% |
Which secrets they use
| Secret type | Proxies | Share |
|---|---|---|
| Plain | 114 | 57% |
| dd (random padding) | 76 | 38% |
| FakeTLS (ee) | 10 | 5% |
Plain secrets still make up the majority, padded dd secrets most of the rest, and FakeTLS stays rare. FakeTLS needs more setup, a believable domain and ideally port 443; 6 of the 10 FakeTLS proxies did run on 443. The domains they imitated, from ya.ru to storage.googleapis.com, are listed in plain, dd and FakeTLS explained.
Names built to be replaced
39 of the 200 proxies were published under a domain name rather than an IP address. In 27 of them the name carried an “ir” label, and many were random subdomains of the same few parent domains. A random subdomain costs nothing, so when one name lands on a block list, the operator publishes a new one pointing at the same server. 34 of the 39 named proxies ran in Germany, which makes them German servers dressed for an Iranian audience.
How long they keep working
Every proxy in the top 200 had answered at least 90% of our checks, and 61.5% had never failed one. That is what the ranking selects for, so this is the best of what circulates, not the average. If yours stops working, switch to a fresh one: proxy not connecting.
How we measured
Our checker collects MTProto proxies from public Telegram channels, tests each one repeatedly and ranks them by how reliably they connect. We took the 200 best-ranked proxies on 9 October 2026, resolved domain names to IP addresses and looked up each address's network and country in public routing data. This is a snapshot of the proxies that work best, not of every proxy in circulation, and response times from our checker are not comparable between proxies, so we do not report them. The live list is on the proxy page.