You've got LM Studio running, a model loaded, the server started, and LMSA still throws "Connection Failed" or just spins forever. It's one of the most common hiccups people hit when setting up local AI on Android, and the good news is that it almost always comes down to one of a small handful of causes. Work through these in order and you'll almost certainly find the culprit before you get to the bottom of the list.
1. Enable "Serve on Local Network" in LM Studio
This is the fix that trips up more people than anything else on this list, mostly because it's easy to miss entirely. Starting the local server in LM Studio is not the same as making it reachable from other devices. By default, the server only listens on your computer itself. In LM Studio's server settings, you need to specifically enable "Serve on local network" after starting the server, this is what actually opens the server up to your phone (or any other device on your Wi-Fi). If this toggle is off, no amount of firewall or IP troubleshooting further down this list will fix anything, because your computer simply isn't listening for outside connections yet.
2. Confirm the Server Is Actually Running
This sounds obvious, but it's still worth a direct check. Loading a model in LM Studio does not automatically start the local server, they're two separate steps. Open LM Studio, go to the Developer (or Local Server) tab in the left sidebar, and check that the toggle actually shows the server as running, not just that a model is loaded. If the model shows as loaded but the server status is off, start it, confirm "Serve on local network" is also on (see #1), and try connecting from LMSA again.
3. Make Sure Both Devices Are on the Same Wi-Fi Network
LMSA talks to LM Studio over your local network, so your phone and your computer both need to be on the same Wi-Fi. This trips people up more than it should, mainly because of two things worth checking specifically:
- Guest networks: Many home routers isolate guest Wi-Fi from the main network by design, so if your phone auto-connected to a guest SSID, it physically cannot reach your computer even though both show "connected" to Wi-Fi.
- 5GHz vs 2.4GHz bands: Some routers broadcast separate SSIDs for each band. Double-check your phone and PC are on the same one, or the same combined network name if your router uses band steering.
If you're not on the same Wi-Fi and want to connect anyway (like when you're out of the house), that's a different setup entirely, using Tailscale, which is covered further down.
4. Use the Right IP Address, Not the One LM Studio Shows You
This is a sneaky one. LM Studio's server tab displays an IP address for you to use, but on Windows machines in particular, that displayed address is sometimes wrong or points to the wrong network adapter (a VPN adapter, a virtual adapter, or a secondary network card), especially on systems with multiple network interfaces.
Instead of trusting what LM Studio shows, pull your actual local IP directly from the OS:
- Windows: Open Command Prompt and run
ipconfig. Look for the IPv4 address under your active Wi-Fi adapter (usually something like192.168.1.XX). - Mac: Go to System Preferences (or System Settings) → Network, select your active connection, and read the IP address shown there.
Enter that address, along with port 1234 (LM Studio's default), into LMSA's connection settings. It should look like http://192.168.1.XX:1234.
5. Check Your Firewall Isn't Blocking LM Studio
Windows Firewall and third-party security software commonly block incoming connections to apps by default, and LM Studio's local server is exactly the kind of thing they flag. If you dismissed or denied a firewall permission prompt when you first started the server, that block sticks around silently.
On Windows: Windows Security → Firewall & network protection → "Allow an app through firewall," then confirm LM Studio is listed and checked for Private networks (you generally don't need Public checked, and shouldn't enable it, for home use). Restart the LM Studio server after changing this.
On Mac, check System Settings → Network → Firewall, and make sure LM Studio isn't in a blocked list.
6. Restart Both Apps
It's not a glamorous fix, but it resolves more connection issues than it should. If LMSA was working before and suddenly stopped, or if you've changed any network, server, or firewall settings, fully close LM Studio (not just minimize it), restart the server, re-enable "Serve on local network," and force-close and reopen LMSA on your phone. Half-restarted states on either side are a common source of connections that look fine but silently fail.
7. Temporarily Disable VPNs
If either your phone or your computer is running a VPN, it can quietly break local network discovery, even if the VPN app claims to allow LAN traffic. VPNs frequently reroute all traffic through a tunnel that doesn't include your home network, which means your phone technically can't "see" your computer's local IP anymore even though both are online. Disable any active VPN on both devices, confirm the connection works, and then decide whether you need the VPN running alongside your local AI setup (if you do, look into split-tunneling options in your VPN app instead of running it globally).
8. Double-Check the Port Number
LM Studio defaults to port 1234, but it's easy to accidentally change this in server settings or forget which port you set previously if you've adjusted it before. Open LM Studio's server settings and confirm the exact port number shown there matches exactly what's entered in LMSA. A single typo'd digit here is an easy miss.
9. Check for IP Address Changes
Most home networks assign IP addresses dynamically (via DHCP), which means your computer's local IP can change after a restart, especially if it's been off the network for a while or your router's lease time is short. If LMSA was working fine yesterday and suddenly can't connect today, re-check your computer's current IP with ipconfig (Windows) or in Network settings (Mac) rather than assuming the old address is still correct. If this keeps happening, most routers let you assign a static/reserved IP to a specific device in their admin settings, which permanently solves the recurring reconnection hassle.
10. Load Models From LMSA, Not From LM Studio Directly
This one isn't strictly a connection failure, but it causes enough "LMSA is broken" reports that it belongs on this list. Once your server is up and reachable, load your model from inside LMSA rather than picking it in LM Studio's own interface. Loading models directly in LM Studio while also selecting one in LMSA can cause multiple models to load and stack in memory at once, which eats up RAM and VRAM fast and leads to sluggish responses, timeouts, or the app appearing to hang, symptoms that look a lot like a connection problem but are actually a resource problem. Get in the habit of picking your model from LMSA's own model selector every time.
If You're Still Stuck: Confirm From the LM Studio Side First
One diagnostic trick worth doing before repeating any of the steps above: open a browser on the same computer running LM Studio and go to http://localhost:1234/v1/models. If that doesn't return a response, the problem is on the LM Studio side (server not actually running, port conflict, etc.) and has nothing to do with LMSA or your phone at all. If it does return a model list, the server itself is healthy and the issue is specifically in the network path between your phone and computer, which points you back to the "Serve on local network" toggle, Wi-Fi network matching, IP address accuracy, or firewall rules.
Connecting From Outside Your Home Network
Everything above assumes you're trying to connect over the same Wi-Fi. If you want to reach your LM Studio server from outside your home (at work, on mobile data, traveling), same-network troubleshooting won't help, because you need a different kind of connection entirely. LMSA has a dedicated Tailscale setup guide in its support documentation for exactly this: it creates a private, encrypted network between your devices that works from anywhere with an internet connection, without port forwarding or exposing your server to the open internet. That's a separate setup process from the local Wi-Fi steps above, but it's worth knowing that "won't connect" when you're away from home isn't a bug, it's just a different setup you haven't done yet.
One more thing worth keeping in mind while troubleshooting: connections over local Wi-Fi between LMSA and LM Studio are unencrypted by design, since it's a direct connection on your own private network. That's expected behavior, not a bug, but it's a good reason to stick to trusted home or personal networks for this kind of setup rather than public Wi-Fi.
Most "won't connect" issues trace back to one of a few things: the local network toggle not enabled in LM Studio, the server not actually started, the wrong IP address, or a firewall quietly blocking the connection. Work through the list in order, and you should have LMSA talking to LM Studio again well before you reach the bottom.