#Fresh Install not working - Claude AI is stumped - please help!

1 messages · Page 1 of 1 (latest)

glossy mistBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> âś… Mark Solution

twilit oceanBOT
#

Two things I’d chase here, because “sometimes first request works, then everything times out” is very consistent with either (a) an HTTP client/runtime issue, or (b) we’re not actually seeing the underlying HTTP error because logging is too low.

1) First: please paste these (they redact secrets)

From the gateway host (same Mac):

openclaw --version
openclaw status --all
openclaw models status --json
openclaw gateway status
openclaw doctor

Those usually make it obvious whether the daemon is using the same config/auth as your shell, and whether provider probes are timing out too.

Docs runbook for this ladder: https://docs.openclaw.ai/gateway/troubleshooting

#

2) Node 25 is a red flag (even if it “should” work)

Docs only say “Node >= 22”, but in practice Node 22 LTS is the most boring/stable choice for long-lived daemons + streaming clients. I would absolutely try running the gateway under Node 22.

Relevant doc line: https://docs.openclaw.ai/help/faq#what-runtime-do-i-need

If you’re using nvm/fnm/mise: on macOS, launchd can keep pointing at a different Node/global install than your interactive shell, so the “I tested with node -e fetch” check can be misleading. The openclaw status --all output should show what the service is actually running.

3) Turn on real HTTP logging for one repro

Right now you only see “LLM request timed out”, which usually means we’re missing the HTTP-layer clue.

For a one-off repro, run the gateway with debug logging + all diagnostics flags (just temporarily):

openclaw gateway stop
OPENCLAW_LOG_LEVEL=debug OPENCLAW_DIAGNOSTICS=* openclaw gateway run