You want a resident personal AI assistant on your own machine: connected to chat channels (Telegram/WhatsApp/Slack/Discord…), able to run scheduled jobs (a 9am daily summary), drive a browser, and remember across sessions — instead of opening a web chat every time. OpenClaw is exactly that: a local Gateway (control plane) + CLI/TUI/web console + channel connectors + scheduled automations (cron) in one personal AI assistant platform, GitHub 388k stars (top open-source project in 2026).

Use it when:

• You want a resident local AI assistant you can remote-control from chat apps (multi-channel);
• You want scheduled/trigger-based automation (daily tasks, periodic summaries, reminders) — openclaw cron is built in;
• You want AI to operate the browser (form filling, posting, cross-page tasks) — openclaw browser is built in;
• You want a skills/plugin ecosystem (ClawHub 54k+ skills) to extend the assistant.

Don't use it when: you only need a one-off conversation (use a chat client); you only need a single-command code tool (that's CLI-tool territory); you are sensitive to resident services/Node runtime (it is a Node platform + resident Gateway).

Zero-account red line: onboard/channel integration needs a model API key or chat account; this guide covers install and CLI verification only, touching no accounts.

One command to install and verify (measured China path: Tsinghua npm mirror prefix):

# Source: Trove detailed guide · verified 2026-09-01
npm_config_registry=https://registry.npmmirror.com npm install -g openclaw@latest --no-audit --no-fund && openclaw --version

Expected output: OpenClaw 2026.8.1 (ea80657) (version follows latest; measured 331 packages in 9s)

Prerequisite: Node ≥ 22.22.3 (or 24.15.0+ / 25.9.0+). If the system Node does not qualify, do not touch it — use an isolated deployment (see "Install variants").

Confirm:

• openclaw --version prints a version;
• openclaw --help lists the full command surface (cron/browser/channels/gateway all present);
• openclaw gateway status exits 0 and gives diagnostics (config missing before onboard is normal).

Install variants (deep dive)

OpenClaw uses a Gateway architecture: one resident Gateway process is the control plane (sessions, tools, events, channel connections); CLI/TUI/web Control UI all connect to it; Channels connect the assistant to WhatsApp/Telegram/Slack/Discord etc.; cron/automations subcommands run scheduled tasks on the Gateway; browser subcommands manage a dedicated browser instance. The model layer supports hosted and local providers (ollama).

Where it writes: npm global package (Node runtime) + first openclaw onboard creates ~/.openclaw/ workspace (config + database + logs) + optional systemd user service (daemon).

Credential behavior (measured key): cron/browser/channels subcommands run through the local Gateway; without onboard they fail with explicit errors instead of silently failing — this is the normal "not configured yet" prompt, not a fault (see section 5 for real output).

vs moltis: OpenClaw = Node-based complete personal AI assistant (multi-channel + cron + browser + skills ecosystem); moltis = Rust single-binary lightweight agent server (security-first). Adjacent positioning, different form: choose by "complete assistant vs lightweight server".

Step 1: check the Node version (prerequisite)

Why this step: openclaw has a hard Node floor (22.22.3+); otherwise npm postinstall refuses (measured: [openclaw] detected Node 22.22.1). Don't touch the system Node — use an isolated deployment.

# Source: Trove detailed guide · verified 2026-09-01
node --version

Expected output: 22.22.3 or higher (22.22.3+ / 24.15.0+ / 25.9.0+ any qualifies)

If stuck:

• node: command not found → install Node first (apt/nvm/official binary);
• version <22.22.3 → deploy an isolated qualifying Node (measured path /opt/node-v24.15.0-linux-x64/bin prepended to PATH), leave the system alone.

Step 2: install (Tsinghua mirror)

# Source: Trove detailed guide · verified 2026-09-01
npm_config_registry=https://registry.npmmirror.com npm install -g openclaw@latest --no-audit --no-fund

Expected output: changed 331 packages in 9s (package ~204MB unpacked; 9s here is cache-accelerated, first install 1-3 minutes is normal)

If stuck:

• [openclaw] detected Node xx → Node version insufficient; back to step 1 isolated deployment;
• enoent spawn sh → PATH polluted (check PATH includes /usr/bin, /bin);
• timeout → if the mirror is unreachable, switch to registry.npmjs.org direct (network dependent).

Step 3: verify the install

# Source: Trove detailed guide · verified 2026-09-01
openclaw --version && openclaw --help | head -20

Expected output: first line OpenClaw 2026.8.1 (ea80657), then the command surface (acp/agent/channels/cron/gateway/memory/models/browser...)

If stuck:

• openclaw: command not found → global bin not on PATH (npm prefix -g to find the global dir; prepend its bin when using an isolated Node).

Step 4: check Gateway status (diagnostics only, no service start)

# Source: Trove detailed guide · verified 2026-09-01
openclaw gateway status

Expected output: Service: systemd user (disabled) + Gateway: bind=loopback (127.0.0.1), port=18789 + Dashboard: http://127.0.0.1:18789/. Config missing before onboard is normal — first full use requires openclaw onboard (interactive wizard, needs a model API key; not covered here).

If stuck:

• config missing → not a fault. Normal before onboard; onboard creates ~/.openclaw/openclaw.json and configures the Gateway.

Step 5: recognize the real pre-onboard errors (common to cron/browser/channels)

Explanation: cron/browser/channels subcommands all go through the Gateway; without credentials they fail with explicit errors. Learn to recognize these so you don't misdiagnose a fault.

# Source: Trove detailed guide · verified 2026-09-01
openclaw cron list

Expected output: gateway cron.list requires credentials before opening a websocket Fix: configure gateway.auth token/password, pair this device, or pass --token/--password. Config: ~/.openclaw/openclaw.json

Meaning: not onboarded / device not paired. Configure gateway.auth (token/password) or pair the device per the hint; if you don't want credentials yet, run openclaw doctor for global status.

# Source: Trove detailed guide · verified 2026-09-01
openclaw browser doctor

Expected output: FAIL gateway: GatewayCredentialsRequiredError: gateway browser.request requires credentials before opening a websocket Fix: configure gateway.auth token/password, pair this device, or pass --token/--password. Config: ~/.openclaw/openclaw.json

Same meaning: browser subcommands all go through the Gateway; pre-onboard they uniformly report the credentials error.

If stuck:

• credentials error looks like a broken install → it isn't. It is the normal "not onboarded yet" prompt; follow the Fix hint;
• want global status first → openclaw doctor health report (model/gateway/channel config; unconfigured items pre-onboard are normal diagnostics).

Step 6: command surface overview (usable after onboard)

Explanation: the following surfaces are visible in --help/command surface (measured); real execution needs Gateway credentials (after onboard). Listed for reference; parameter details in official docs docs.openclaw.ai.

• cron/automations: add new, list list, show/get view one, edit edit, enable/disable, rm delete, run run once, runs history, status scheduler state, scratch private scratchpad;
• browser: create-profile/delete-profile manage browser profiles, click/click-coords/fill/drag/highlight page interaction, evaluate run a function on a page, close close tab, console/errors page logs, download save downloads, batch run a batch of actions, doctor readiness;
• channels: add add/update channel, login/logout link/unlink sessions, remove delete, status --probe probe, logs channel logs, capabilities provider abilities. Non-interactive example (needs a real token): openclaw channels add --channel telegram --token <token>.

Same source as the tool-page matrix; full pairs, states and evidence on the conflict matrix. Known relations:

• moltis × openclaw: adjacent positioning (full assistant vs lightweight server), different command names, non-overlapping write surfaces (each ~/.config dir), static: can coexist;
• herdr/ruflo/claude-mem etc.: non-overlapping write surfaces, static: can coexist;
• Others: per the matrix page (mostly possible-compatible inferred).

• Symptom: npm reports [openclaw] detected Node 22.22.1 → system Node below 22.22.3+; deploy an isolated qualifying Node (measured /opt/node-v24.15.0-linux-x64), don't touch the system Node
• Symptom: gateway status shows config missing → not a fault; normal before onboard; onboard creates the config
• Symptom: cron/browser report credentials required → not onboarded/device not paired; configure gateway.auth or pair per the Fix hint
• Symptom: enoent spawn sh → PATH polluted (missing /usr/bin, /bin)
• Symptom: openclaw command not found after install → global bin not on PATH (npm prefix -g; prepend bin for isolated Node)

Full four-section entries (symptom→cause→fix→evidence): site docs (if recorded)

# Source: Trove detailed guide · verified 2026-09-01
npm uninstall -g openclaw --no-audit --no-fund && rm -rf ~/.openclaw

Expected output: npm removal output (Removed N packages); ~/.openclaw gone = clean

Note: deleting the workspace loses all config/memory/channel bindings; run openclaw backup first if you want to keep them; stop the daemon first if enabled.

• Q: npm reports [openclaw] detected Node 22.22.1 — what now? A: system Node is below 22.22.3+. Don't touch it — deploy an isolated qualifying Node (e.g. 24.15.0) and install with its PATH prefix (measured path: /opt/node-v24.15.0-linux-x64/bin). The openclaw command then also uses that Node prefix.
• Q: gateway status shows config missing — is it broken? A: No. Normal before onboard; onboard creates ~/.openclaw/openclaw.json and configures the Gateway.
• Q: openclaw command not found after install? A: check the global bin on PATH (npm prefix -g); with an isolated Node, make sure its bin is first in PATH.
• Q: Does onboard really need an API key? Can I use local models? A: officially both hosted and local providers are supported (ollama etc.). onboard is an interactive wizard; not tested here (zero-account); details at docs.openclaw.ai/concepts/model-providers.
• Q: Do scheduled tasks need the Gateway running all the time? A: Yes — cron/automations run on the Gateway and need the daemon resident. Not tested (zero-account + avoiding resident processes), honestly marked.
• Q: What's the difference between OpenClaw and moltis? A: OpenClaw = Node-based complete personal AI assistant (multi-channel + cron + browser + skills ecosystem); moltis = Rust single-binary lightweight agent server (security-first). Adjacent but different form: choose by "complete assistant vs lightweight server".
• Q: Where do I report issues? A: install/uninstall issues go through this entry's feedback; project issues go to the author repo.

• Version: 2026.8.1 (manifest locked; npm measured 2026-09-01).
• Environment: Node ≥22.22.3 (24.15.0+/25.9.0+ OK); isolated deployment measured at /opt/node-v24.15.0-linux-x64.
• Network: npm via Tsinghua mirror (measured 331 pkgs/9s); official registry direct is slow.
• Write surface: ~/.openclaw/ (config/database/logs) + npm global package; optional systemd user service.
• Port: Gateway 18789 (loopback); onboard/channel integration needs credentials (zero-account, not covered).

• Sources: official project docs (docs.openclaw.ai) + local measurements (2026-09-01)
• Entry signature: openclaw official npm package (kept as 2026.8.1)

• Measured: local repro log 2026-09-01 (Node 24.15.0 isolated deploy npmmirror 31MB/10s + Tsinghua mirror npm install 331 pkgs/9s + openclaw --version 2026.8.1 + --help surface + gateway status loopback 18789 exit 0)
• Not tested: onboard wizard (needs API key, zero-account), resident Gateway service, channel integration, actual cron/browser execution (surface visible but needs onboard).