ComfyUI Reconnecting Error: Causes and Fast Fixes

If ComfyUI keeps showing "Reconnecting...", the browser has lost its live connection to the ComfyUI server.

That does not always mean your workflow is broken. It can mean the server crashed, restarted, froze under load, or the frontend websocket connection was blocked. The fastest path is to find out whether ComfyUI is still running in the background.

Quick Diagnosis

Step 1: Check Whether the Server Is Still Alive

Open the ComfyUI terminal window. If the terminal is gone, frozen, or shows a traceback, treat the reconnecting message as a server-side failure.

If the terminal is still running, try opening:

http://127.0.0.1:8188

Use 127.0.0.1 before testing a LAN IP, reverse proxy, or remote address. This separates local ComfyUI problems from network access problems.

If you changed the port, use that port instead of 8188.

Step 2: Read the First Real Error in the Terminal

Do not start with the last line. Scroll up and look for the first serious error:

Traceback (most recent call last):
ModuleNotFoundError
ImportError
CUDA out of memory
RuntimeError
ConnectionResetError

Everything after the first error may be a secondary symptom. For example, a plugin import failure can break the server, and the browser only reports the visible reconnecting state.

The official ComfyUI troubleshooting guide also recommends collecting terminal logs, browser console errors, installed custom nodes, and reproduction steps before reporting a bug.

Step 3: Test Without Custom Nodes

Custom nodes are a common reason ComfyUI starts reconnecting after an update or plugin install.

For a manual install, run:

python main.py --disable-all-custom-nodes

For the Windows portable package, create a copy of your launch batch file and add:

--disable-all-custom-nodes

If ComfyUI works with custom nodes disabled, the core install is probably fine. The problem is likely one custom node, one frontend extension, or a dependency installed by a custom node.

Then narrow it down:

1. Re-enable half of the custom nodes.
2. Restart ComfyUI.
3. If reconnecting returns, the bad node is in that half.
4. Repeat until you find the plugin.

For deeper plugin failures, use How to Fix ComfyUI Plugin Import Failed Errors.

Step 4: Update Frontend and Requirements Together

Some reconnecting problems happen after updating only part of the stack.

If you update ComfyUI with git pull, also update dependencies in the same environment:

pip install -r requirements.txt

If you use the portable package, use the bundled Python, not a random system Python. A common mistake is installing packages into the wrong environment and then wondering why ComfyUI still fails.

For update-related dependency problems, see ComfyUI Dependency Conflicts.

Step 5: Check Browser and Websocket Issues

If the server is alive but the browser keeps reconnecting, check the browser side.

Try:

• reload the page with Ctrl+F5
• use Chrome or Edge with extensions disabled
• open DevTools, then check the Console and Network tabs
• test http://127.0.0.1:8188 instead of a LAN or proxy address
• temporarily disable VPN, proxy, or aggressive antivirus web filtering

ComfyUI relies on a live browser-to-server connection. Browser extensions, proxy tools, remote tunnels, and corporate security software can interrupt that connection even when the Python server is healthy.

Step 6: Check Port and Firewall Problems

If another app is using port 8188, ComfyUI may fail to bind correctly or you may connect to the wrong process.

On Windows, check the port:

netstat -ano | findstr :8188

If the port is busy, start ComfyUI on another port:

python main.py --port 8189

Then open:

http://127.0.0.1:8189

If you access ComfyUI from another device, make sure the firewall allows the port and that ComfyUI is actually listening on the right network interface.

Step 7: Check GPU or Memory Crashes During Generation

If reconnecting appears only after pressing Queue or while a workflow is running, check for runtime crashes:

• CUDA out of memory
• driver reset
• model architecture mismatch
• custom node execution error
• Python process killed by the OS

Try a smaller workflow:

• lower resolution
• reduce batch size
• close other GPU-heavy apps
• test with the default workflow
• run with --lowvram if your GPU is tight

If the terminal shows CUDA out of memory, see Python Out of Memory in ComfyUI and GPU Compatibility.

When Reinstalling Is the Wrong First Move

Reinstalling can hide the real cause and cost you time if you already have:

• downloaded models
• custom nodes
• working workflows
• project-specific settings
• client or production assets

Use reinstall only after you know whether the problem is core ComfyUI, a custom node, a dependency conflict, or a network/browser issue.

How Wonderful Launcher Helps

Wonderful Launcher is designed for the phase where a ComfyUI environment is no longer a clean install:

• plugins changed dependencies
• startup logs contain multiple failures
• a custom node update broke the frontend
• you need to preserve models and workflows
• you want a clearer recovery path than random shell commands

If reconnecting keeps returning after custom node or dependency work, start with Wonderful Launcher before wiping the environment.

What to Send When Asking for Help

Send these details:

Related Guides

• ComfyUI Failed to Fetch Server Logs
• ComfyUI Startup Failed? How to Diagnose and Recover Faster
• How to Fix ComfyUI Plugin Import Failed Errors
• ComfyUI Dependency Conflicts
• ComfyUI Common Issues and Fast Fixes

Source References

• Official ComfyUI troubleshooting overview
• Official custom node troubleshooting guide
• Official ComfyUI update troubleshooting