VSCO and Proxies: Scope Difference, Uploading and Shared Access

When you look at VSCO from the network side, you see two different behaviours. The phone app does the editing work on the device's own processor and goes out to the network only with session, upload, feed and profile requests. The vsco.co profile opened in a browser is a classic web client: document, script and image requests each set off separately.

Setups that skip this distinction produce problems that look alike but have different sources. When you define a proxy on the phone and then switch to mobile data, the rule quietly stops applying. When you use a rule in the browser that covers only the main domain, the page opens but the images stay blank. Both look like "the proxy isn't working"; yet one is an interface scope problem and the other a domain scope problem.

The sections below cover first where the traffic is distributed, then the scope difference between the two clients, the upload pipeline, the real-time channel and the access arrangement in team use.

Which endpoints does VSCO traffic go to?

When you open a session, the client first performs a TLS handshake with the authentication endpoint; the address the other side sees is the proxy server's exit address. The client then carries a session token. The token's validity and the network the connection comes from are evaluated separately: even if the token is correct, additional verification may be requested if the origin of the connection looks unusual.

The second group is the discover feed and profile galleries. Preview images generally come from a separate media domain, and these are the most numerous group of requests. The third group is upload and publish requests; because they carry a body, they are the part most sensitive to connection quality.

The three groups are not equally sensitive to the proxy either. Authentication requests are small but critical: a drop here breaks the session entirely. Media requests are high in volume and, even if individual ones fail, only produce missing images. Upload requests are both high in volume and critical, because a transfer that is cut short starts over.

When writing rules, you need to think about these three groups separately. A rule written for a single domain in practice routes only the first group, and this is the most common reason a setup "looks like it works" while remaining incomplete.

The phone app and the browser profile do not see the same rule

On the mobile side the standard way to define a proxy is to enter an HTTP proxy in the settings of the Wi-Fi network you are connected to. That rule is tied only to that network profile: the moment the phone falls back to mobile data, traffic goes directly over the carrier and you see no warning. For setup steps, the Android proxy settings and iPhone proxy settings articles give a detailed sequence.

There is a second limit as well: some apps never read the system- or network-level proxy setting and open their own connection directly. This behaviour is not a bug but a client design choice. As a result, in a mobile setup it is not enough to say "I applied the setting"; you need to see from an independent source that the exit really changed.

On the browser side the scope is more manageable, but how you set it up depends on the browser: Firefox keeps separate network settings per profile, while Chrome and Edge use the system proxy setting; in those browsers a per-profile exit requires an extension or launching the profile separately with its own proxy parameter. When the setup is done correctly, your other tabs are unaffected and it becomes possible to work with different exits on the same computer. In return, the browser brings two risks the mobile app never produces: name resolution happening locally, and the WebRTC interface exposing your real address.

Verifying both clients with the same checklist is therefore misleading. On the phone the decisive question is "which network interface"; in the browser it is "which domain and which layer".

Editing ends on the device; the network load begins at upload

VSCO's distinguishing feature is that its presets and editing tools run locally. When you open a photo and play with the tones, no request goes to the network; proxy latency therefore does not affect the editing experience. The network load begins the moment you decide to publish the edited file.

This also changes expectations. Because the proxy adds an extra hop, it generally lengthens connection time rather than shortening it. The only places you feel latency are scrolling the feed and the upload stage. For conceptual background see the proxy latency article, and for the measurement side see how to test proxy speed guide.

On the upload side, what matters is not raw speed but the stability of the connection. An exit that offers high bandwidth but drops frequently gives worse results than a narrower but uninterrupted exit, because every drop causes the transfer to be repeated. If you work with high-resolution files, this difference grows quickly.

• Verify your exit with the My IP address tool before uploading.
• With large files, look at connection stability as well as bandwidth.
• Do not run a VPN and a proxy at the same time; it becomes unclear which one is routing.
• Do not run proxy tests on a weak wireless network; the measurement will be distorted by the network.

If an upload stalls, which stop should you look at?

An upload is not a single request. First it is verified that the session is still valid, then the file body is transferred in chunks, and finally the publish record is created. If all three steps do not go over the same proxy, the process can stall halfway: the progress bar fills in the interface but the post does not appear.

Body transfer is the most fragile point in the proxy chain. An idle timeout on the server side can cut off a long-running transfer. Connection pooling and keep-alive behaviour are decisive at this point; the keep-alive and connection pooling article explains how to read timeout settings.

Retry behaviour matters too. Repeating the same request back to back immediately after a drop usually does not fix the situation and generates unnecessary load on the server side. A wait with a progressively increasing interval both yields a higher success rate and creates a calmer pattern on the other side.

If the drops keep recurring, first measure the proxy's liveness with the proxy checker tool, then try the same file on a network without a proxy. The difference between the two attempts quickly separates whether the problem is in the network or in the client.