Where Did Your OpenClaw First Run Stall?
Check Node.js against OpenClaw’s documented version ranges, then record separate times for opening the dashboard and receiving a fresh model reply. Keep those observations with the active config path so you can reproduce a failure.
Record the two milestones on the first run
For a CLI onboarding test, record the Node.js version from the shell you will actually use. The supplied OpenClaw docs list Node 24.16.0 or later within 24.x, or 26.1.0 and later, and give higher patch targets for security fixes on the 24 and 26 lines. Start onboarding with `npx openclaw@latest`. Choose Quick start to try detected AI access, or Custom setup to walk through the guided options. If automatic verification fails, finish the manual provider setup. Start your stopwatch when onboarding begins. Mark the first usable dashboard, then immediately send an original prompt such as “Return the sum of 19 and 23 as digits only.” Record the arrival of the correct reply as the second milestone. A dashboard at 42 seconds and a correct reply at 67 seconds leaves a 25-second interval to investigate, including your time to submit the prompt.
Getting started - OpenClawCheck the active config before repeating setup
If onboarding appears to succeed but a later launch uses unexpected settings, record whether `OPENCLAW_CONFIG_PATH` is set in the failing launch context. OpenClaw documents `~/.openclaw/openclaw.json` as the default, and config writes require the active path to point to a regular file. This matters when an operator moves state onto another disk and leaves a file symlink behind. A readable snapshot can coexist with a failed settings write because snapshot reads can follow a config-file symlink. Use `openclaw config get agents.defaults.workspace` to observe the workspace value, then compare it with the workspace shown in the Control UI’s Config tab and your deployment notes. If a custom path is intentional, set it directly to the real config file in the relevant launch context. Reopen the dashboard and repeat the reply check after confirming the expected workspace.
Configuration - OpenClawRun onboarding from a healthy Gateway, not a clean terminal
Get the Gateway healthy before you touch the wizard. The documented route starts the guided onboarding automatically at install, and macOS, Linux, and WSL2 users can trigger it via the install script. Either way, the Gateway must be listening on port 18789, which you confirm with the gateway status check in the official docs.
If the Gateway is not up, onboarding prompts will run against nothing. Stop the foreground process, install the service, then run the start command, which waits until health and readiness checks pass. Only after status shows the listener should you re-run the onboarding wizard; the documented flag that opens the classic step-by-step flow is the one to use when you want every guided option exposed.
This ordering matters because the config is saved between the stop and the service install. You are not losing provider settings by restarting the Gateway first.
Getting started - OpenClawPick Quick start when existing AI access already works
The wizard's first real fork is Quick start versus Custom setup. Choose Quick start if you already have a Claude Code or Codex CLI login or a provider API key on the machine; the docs state OpenClaw finds that access, verifies it with a real completion, saves the config, and opens the web dashboard. That verification is your first observable success signal: a completion actually returned, not just a key that was stored.
Choose Custom setup when you want to walk every guided option, including provider sign-in and optional settings. If no detected route works, onboarding falls back to manual provider setup on its own, so a failed detection is not a dead end. Provider sign-in and optional setup can take longer than the quick path; you can return later with the configure command for additional settings rather than restarting the whole wizard.
Getting started - OpenClawKeep it in the background only if the session must outlive the terminal
Quick start asks whether to keep OpenClaw running in the background. Answering yes installs a copy of the same version under the home directory config folder and runs the Gateway as a background service, so clearing the npx cache does not stop it. Answering no keeps the Gateway in the terminal until you press Ctrl+C, while the config stays saved either way.
Practical consequence: if you answer no, close the terminal and the agent stops answering even though onboarding succeeded. That looks like a broken provider but is a stopped process. Re-run the gateway status check before debugging credentials. The service install is registered before the Gateway is ready, so pair it with a start command that waits for health rather than assuming registration equals readiness.
Getting started - OpenClaw