Browser tool works sometimes but times out randomly. Chromium headless running as systemd service on port 18800, confirmed working (curl http://127.0.0.1:18800/json/version returns Chrome/147.0.7727.116). OpenClaw browser doctor shows all OK. But when agent tries to use browser tool via Telegram, it times out intermittently. Lisa Romano website worked once, YouTube channel always times out. Is there a way to connect OpenClaw browser plugin to an already-running Chromium on port 18800 instead of launching its own? Config has browser.executablePath set. OpenClaw v2026.5.3-1, Hostinger VPS.
#Browser tool intermittent timeout - how to connect to existing Chromium instance?
1 messages · Page 1 of 1 (latest)
Yes — browser.executablePath only tells OpenClaw what binary to launch. To use your already-running Chromium, switch browser control into CDP attach mode.
Try this shape in ~/.openclaw/openclaw.json:
{
browser: {
enabled: true,
defaultProfile: "openclaw",
attachOnly: true,
cdpUrl: "http://127.0.0.1:18800",
headless: true,
noSandbox: true,
remoteCdpTimeoutMs: 5000,
remoteCdpHandshakeTimeoutMs: 10000,
ssrfPolicy: {
allowedHostnames: ["127.0.0.1", "localhost"],
hostnameAllowlist: ["127.0.0.1", "localhost"]
}
}
}
Then restart the gateway and smoke-test the same path the agent will use:
openclaw gateway restart
openclaw browser --browser-profile openclaw start
openclaw browser --browser-profile openclaw status
openclaw browser --browser-profile openclaw tabs
openclaw browser --browser-profile openclaw open https://example.com
Docs: https://docs.openclaw.ai/tools/browser#local-vs-remote-control and Linux browser troubleshooting: https://docs.openclaw.ai/tools/browser-linux-troubleshooting
Two notes:
- If your Chromium is already on
18800,attachOnly: trueis the key part. Without it, OpenClaw may still try to manage/launch its own browser. - YouTube timing out while simpler sites work may be separate from CDP attachment — YouTube is heavy and often rough on headless VPS browsers. If attach mode works for
example.com/normal sites but YouTube still hangs, grabopenclaw logs --followduring the attempt and check whether it’s CDP connect timeout vs page navigation/load timeout.
openclaw browser start times out after 30000ms even though localLaunchTimeoutMs is set to 60000. Config has proper profile with cdpPort 18800. Chromium works manually (chromium-browser --headless --no-sandbox --remote-debugging-port=18800 starts fine). Is there a separate CLI timeout for browser start command? Or should I use attachOnly with manually started Chromium instead?
Yep — there are two different timeouts here.
browser.localLaunchTimeoutMs is the Gateway’s internal Chrome/CDP startup budget. The CLI also has its own request timeout, and openclaw browser defaults to 30000ms unless you override it.
Try:
openclaw browser --timeout 90000 --browser-profile openclaw start
or, if you want to test the raw control service:
curl -s -X POST http://127.0.0.1:18791/start
That said: on a Hostinger VPS, if manually started Chromium is reliable, I’d use attachOnly. It avoids OpenClaw repeatedly spawning/monitoring Chromium, which is often the flaky bit on low-end/headless Linux.
Recommended shape:
{
browser: {
enabled: true,
defaultProfile: "openclaw",
attachOnly: true,
cdpUrl: "http://127.0.0.1:18800",
remoteCdpTimeoutMs: 5000,
remoteCdpHandshakeTimeoutMs: 10000,
ssrfPolicy: {
allowedHostnames: ["127.0.0.1", "localhost"],
hostnameAllowlist: ["127.0.0.1", "localhost"]
}
}