openclaw gateway status shows offline right after onboard

Re-run openclaw doctor, confirm the daemon unit is loaded, verify listen ports are not blocked by another process, and ensure API tokens in the gateway config match the workspace you onboarded.

Should budget teams use install.sh or npm -g for OpenClaw?

Prefer install.sh when you want a verified curl pipeline and fewer global npm surprises on a fresh lease; prefer npm -g when you already pin Node twenty-four with nvm and commit package-lock style discipline in your own repo.

What disk watermark should trigger a storage upgrade on M4 sixteen gigabytes?

Treat thirty-five percent free space during Skills cold start as a hard pause signal; sustained peaks above forty-five gigabytes of writable cache on a two hundred fifty-six gigabyte SSD mean you should read the storage matrix and consider two hundred fifty-six or five hundred twelve gigabyte tiers.

2026 vuzcloud Budget OpenClaw Install Matrix: install.sh vs npm, Node 24 & M4 16GB

Budget founders need a low-risk OpenClaw install path on a rented Mac mini M4 before they buy metal. Skim home, open pricing for starter tiers, and read help for SSH—then use this matrix for install.sh versus npm -g, Node 24, and M4 sixteen gigabyte peak checks. Pair it with the lease RTT matrix when you pick a region.

01install.sh versus npm -g: OpenClaw install path comparison

Treat each row as a field heuristic on a fresh vuzcloud lease, not a vendor SLA. Confirm node -v reports major version twenty-four before either path touches global binaries.

Decision axis	Official install.sh (curl)	npm install -g openclaw	Budget-team cue
Reproducibility	Pinned script + digest check	Depends on your lock discipline	Prefer install.sh on shared student leases
Node 24 gate	Script aborts early	Works if nvm default is 24.x	Run `node -p process.versions` first
Upgrade cadence	Re-run curl when release notes ship	npm -g update openclaw	Pin one path per sprint; never mix mid-onboard
Disk on 256GB M4	Lean global footprint	npm cache can spike during cold install	Clear ~/.npm after failed attempts

After install, record the CLI hash with which openclaw and openclaw --version in your team wiki so the next teammate does not accidentally run a second global copy.

02openclaw onboard --install-daemon acceptance on a leased Mac

Onboard is your contract test: gateway config, workspace tokens, and a daemon that survives SSH disconnect. Run these checks in order; stop when any step fails—do not stack Skills on a broken gateway.

Preflight Node: node -v must show v24.x; if not, switch with nvm before onboard
Onboard + daemon: openclaw onboard --install-daemon and confirm LaunchAgent (macOS) shows loaded
Gateway listen: openclaw gateway status reports online with expected port
Doctor pass: openclaw doctor exits zero; archive stderr if it does not
Token scope: API keys are workspace-scoped; rotate if you reused a classroom key on a shared lease

Lease hygiene snapshot ~/.openclaw paths you will wipe on return; never leave production tokens on a daily tier host past the spike window

03Skills cold start: disk and memory peak thresholds on M4 16GB

Skills imports are where budget leases fail silently: npm tarballs, model caches, and gateway buffers compete on the same two hundred fifty-six gigabyte SSD. Measure once with df -h / and Activity Monitor before you declare victory.

Signal	Pause threshold (16GB / 256GB)	Upgrade cue
Disk free	Below 35% free during cold import	See storage matrix for 256 vs 512GB
RSS peak	Sustained above 12 GB with GUI open	Close VNC; run headless gateway only
Writable cache	Above ~45 GB cumulative caches	Prune Skills vendor dirs; extend lease tier

For Transporter-heavy workflows after OpenClaw stabilizes, read the US West Transporter runbook—it assumes the same Node and disk guardrails.

04US West versus APAC: pick a vuzcloud node for OpenClaw latency

Install path does not fix geography. Pick US West when your model APIs, webhooks, and git remotes sit in North America. Pick APAC when commits and reviewers are in East Asia—even if some US APIs add RTT.

Quantify with the APAC vs US West RTT matrix and sample three endpoints you actually call from openclaw gateway status, not a single ping chart.

05Pain signals that waste a budget OpenClaw spike

Node drift: Homebrew Node twenty-two coexists with nvm twenty-four; onboard succeeds then gateway crashes on the wrong binary
Dual globals: install.sh plus npm -g leaves two openclaw binaries; doctor flags pass on the wrong one
Disk cliff: Skills cold start fills the SSD while gateway logs still look healthy—free space is the real SLO

06Five-step budget runbook from SSH to first Skill

Step one: SSH in, run node -v, and pin Node 24 with nvm if the lease image is ambiguous.

Step two: Choose install.sh or npm -g from the table; never run both on the same host in one afternoon.

Step three: Execute openclaw onboard --install-daemon; capture gateway status and doctor output in your repo docs.

Step four: Import one representative Skill bundle; watch disk free and RSS against the thresholds above.

Step five: Compare daily versus weekly tiers on pricing using the RTT matrix if the spike lasts more than a few billable days—release the lease when validation ends.

Minimum Node major version for OpenClaw install

35%

Minimum free disk before pausing Skills import

12 GB

RSS pause line on M4 16GB with GUI closed

Citable guardrails: Node 24 is a hard floor; install.sh is the default on shared student leases; npm -g is fine when nvm already pins twenty-four.

Disk guardrail: pause Skills when free space drops below thirty-five percent on a two hundred fifty-six gigabyte SSD.

Related reads: budget storage matrix, blog index, help for SSH and VNC basics.

07FAQ: gateway status, doctor, and install failures

gateway status offline after onboard: Re-run openclaw doctor, reload the daemon unit, and confirm no other process owns the listen port. Verify tokens match the workspace you onboarded.

doctor reports Node mismatch: Print which node and which openclaw; align PATH so both resolve under nvm twenty-four, not Homebrew twenty-two.

npm EACCES on global install: Prefer install.sh on leased Macs, or prefix npm global to a user-owned directory—never sudo on a shared vuzcloud host.

Thresholds summarize typical field practice on leased Apple silicon hosts. Always let checkout and contract pages override illustrative numbers shown in editorial content.

Starter OpenClaw on cloud Mac

Rent an entry Mac mini M4 and run the install matrix today

Read budget storage matrix or Transporter webhook runbook, then open pricing for entry tiers and purchase to provision your first node.

Rent entry M4 now View starter plans