# How to set up OpenClaw on macOS. > Set up OpenClaw on macOS with the official CLI flow: install, onboard, pair a channel, open the dashboard, and run the basic health and security checks. - Canonical page: [https://veridicuscan.app/openclaw-macos-setup](https://veridicuscan.app/openclaw-macos-setup) - Language: English - Published: 2026-03-10 - Last materially updated: 2026-03-10 Practical guide Use the official OpenClaw flow to install the CLI, run onboarding, pair a channel, and verify the gateway on macOS without detouring through unofficial fixes. Official baseline - macOS 11 or later, Intel or Apple Silicon - Node.js 22 or newer for the package-install path - Onboarding flow: install, onboard, channel login, dashboard - Security checks belong in the initial setup, not later Step 1 ## Install OpenClaw with the documented CLI path. The official docs show both an installer script and a standard package install. For a setup post, the explicit CLI path is easier to read and easier to troubleshoot on macOS. MACOS / INSTALL / CLI check node **`node --version`** install **`npm install -g openclaw@latest`** or **`pnpm add -g openclaw@latest`** verify **`openclaw --help`** Step 2 ## Run the onboarding wizard and let it install the background service. OpenClaw’s documented recommended path is the onboarding wizard. It sets up the gateway, workspace, channels, skills, and a user-level background service so the gateway can keep running. 01 ### Run onboarding `openclaw onboard --install-daemon` 02 ### Choose your model and auth The wizard will walk through provider and authentication choices before it drops you into the rest of the setup. 03 ### Keep the generated defaults For a first setup, keep the safer defaults and change them later only after the gateway is working. Step 3 ## Pair a channel and open the dashboard. After onboarding, log in to the chat channel you actually want to use, then open the local dashboard. If you only want the quickest first success, OpenClaw’s docs note that Telegram is usually the fastest channel to set up. ### Pair a channel - `openclaw channels login` - Follow the prompts for the channel you chose - Use the simplest path first, then add more channels later ### Open the dashboard - `openclaw dashboard` - Or open `http://127.0.0.1:18789/` - If auth is required, use the local tokenized dashboard link ### Verify the gateway - `openclaw status --all` - `openclaw doctor` - `openclaw security audit` macOS notes ## Fix the common Mac issues while the setup is still small. The official platform docs call out a few macOS-specific issues that are easier to solve before you start adding more channels, tools, and automation. 01 ### Gatekeeper If macOS blocks the binary, the official docs suggest removing the quarantine flag from the installed executable path. 02 ### Privacy permissions Accessibility, Full Disk Access, Camera, or Microphone prompts depend on the features you enable. Grant only what your workflow actually needs. 03 ### Background service If you change install method or the gateway behaves oddly after an update, run `openclaw doctor` to repair the service entrypoint. Before you turn on risky tools ## Set up OpenClaw first. Then add Veridicus Scan before it reads untrusted content. OpenClaw’s own security docs warn that prompt injection can come from external content, not just from who messages the bot directly. Once the setup works, the next product to add is Veridicus Scan so suspicious pages and files get reviewed locally before they reach a tool-enabled agent. ### Run the security audit Use `openclaw security audit` before you widen access, expose tools, or trust the setup enough to automate it. ### Keep high-risk tools narrow Do not switch on `browser`, `web_fetch`, `web_search`, or `exec` broadly until you understand the blast radius. ### Scan untrusted inputs first Use Veridicus Scan on suspicious URLs, PDFs, DOCX files, and HTML before that content becomes input to a tool-enabled OpenClaw workflow. Next step ## Use OpenClaw for the agent. Use Veridicus Scan for the intake decision. When the agent will read the web or imported files, put a local scan step in front of it so hostile content is caught before it becomes agent context.