ameshkov

Joined 9 years ago

Karma
2,363
Submissions
196
Member since
Oct 15
https://twitter.com/ay_meshkov
View on Hacker News

Recent submissions

Comment · 2 months ago
I made an MCP proxy with a similar idea in the past: replace a ton of tools that consume tokens with just two (get_tool_schema, invoke_tool) - https://github.com/ameshkov/mcp-compress-router

One thing that I noticed is that it’s often better to return tool names with argument names, i.e. return “search_web(query)” instead of just “search_web” when listing tools. Otherwise models often tend to hallucinate argument names and an extra turn is required to correct the mistake.

One additional advantage that such tools provide is that when you use different coding agents you don’t have to set up all the MCP servers in every agent, you just set up one (or point the agent to the cli like in this project).

Comment · 2 months ago
Comment · 8 months ago
Comment · 9 months ago
No, but it's a very good point, we'll add it to the backlog
Comment · 9 months ago
In Cyprus, companies are usually incorporated via a corporate service provider, who also provides the registered office address and PO Box services. Basically, if you need to reach out to us by mail, you should use the "legal address" from the website.
Comment · 9 months ago
For example, in some countries it's either slowed down or outright blocked.
Comment · 9 months ago
Actually, no, we don't resolve them. We scan the incoming ClientHello before making a decision on where to route the connections. If the connection should be bypassed we make a connection by ourselves and proxy traffic. Implementing it that way requires having a TCP stack right in the client.
Comment · 9 months ago
Well, Google is like a small country and some parts of it are very good guys that are genuinely interested in being a "user agent".
Comment · 9 months ago
Uh, I guess it's a little bit off-topic here:) It's hard to say what's more harmful, I'd say cookies still take #1, but I think we're not far from the moment when your email address or its derivative will be used as the main advertising ID. Regarding evolution, well, definitely possible, the time will show.
Comment · 9 months ago
This is actually supported by both the client and the server.

To use it in mobile clients you need to specify two domain names like that: fake-sni.com|domain.com where “fake-sni.com” is the domain thay will be in the SNI and “domain.com” is the domain in your TLS certificate (used to check the server’s authenticity)

Comment · 9 months ago
There are no shady alleys in Cyprus:)

We have only one office in Limassol, the company is mostly remote: https://maps.app.goo.gl/pounSEQqBvYftZGZ6?g_st=ic

(we are moving to a bit bigger office in the neighboring building, no nice photos on google yet)

We do not have a dedicated team page on the website, but we’re not hiding our faces, the team can be found on Github. Members of the team often visit AFDS [1] [2], you can see some faces there (including mine).

[1]: https://adfilteringdevsummit.com/

[2]: https://youtube.com/playlist?list=PL61EKVIQWizG0tIYqNDoenVaO...

Comment · 9 months ago
Comment · 9 months ago
You mean in TrustTunnel apps? You can create a routing profile there and select which domains/ips are bypassed, and then select that routing profile in the vpn connection settings.
Comment · 9 months ago
We support both H2 and H3 and this is necessary. QUIC is not bad, but there are places where it either does not work at all or works too slow.

And one more thing, even though the code and spec is only published now, we’ve been using TrustTunnel for a long time, started before CONNECT_UDP became a thing.

We’re considering switching to it though (or having an option to use it) just to make the server compatible with more clients.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton