openclaw/docs/cli/browser.md

summary	read_when	title
CLI reference for `openclaw browser` (lifecycle, profiles, tabs, actions, state, and debugging)	You use `openclaw browser` and want examples for common tasks You want to control a browser running on another machine via a node host You want to attach to your local signed-in Chrome via Chrome MCP	Browser

openclaw browser

Manage OpenClaw's browser control surface and run browser actions: lifecycle, profiles, tabs, snapshots, screenshots, navigation, input, state emulation, and debugging.

Related: Browser tool

Common flags

• --url <gatewayWsUrl>: Gateway WebSocket URL (defaults to config).
• --token <token>: Gateway token (if required).
• --timeout <ms>: request timeout in ms (default: 30000).
• --expect-final: wait for a final Gateway response.
• --browser-profile <name>: choose a browser profile (default: openclaw, or browser.defaultProfile).
• --json: machine-readable output (where supported). This is a browser-level option, so place it before the subcommand for an unambiguous form, such as openclaw browser --json status. Trailing placement such as openclaw browser status --json also works when the selected child command does not define its own --json.

With --json, local action-input, state, and viewport validation refusals write one JSON error document to stdout (ok: false, error.type: "cli_error") and exit with status 1. Without --json, the diagnostic remains on stderr.

Malformed request timeouts, invalid tab indexes, and Gateway request rejections also use this JSON failure envelope, including navigation, resize, and cookie requests.

The CLI reserves 10 additional seconds beyond the request timeout for node and Gateway transport, allowing browser timeout diagnostics to reach the caller.

Quick start (local)

openclaw browser profiles
openclaw browser --browser-profile openclaw start
openclaw browser --browser-profile openclaw open https://example.com
openclaw browser --browser-profile openclaw snapshot

Agents can run the same readiness check with browser({ action: "doctor" }).

Quick troubleshooting

If start fails with not reachable after start, troubleshoot CDP readiness first. If start and tabs succeed but open or navigate fails, the browser control plane is healthy and the failure is usually a navigation SSRF policy block.

Minimal sequence:

openclaw browser --browser-profile openclaw doctor
openclaw browser --browser-profile openclaw start
openclaw browser --browser-profile openclaw tabs
openclaw browser --browser-profile openclaw open https://example.com

Detailed guidance: Browser troubleshooting

Lifecycle

openclaw browser status
openclaw browser doctor
openclaw browser doctor --deep
openclaw browser start
openclaw browser start --headless
openclaw browser stop
openclaw browser --browser-profile openclaw reset-profile

• doctor --deep adds a live snapshot check: useful when basic CDP readiness is green but you want proof the current tab can be inspected.
• For a running local managed profile, status and doctor report cached graphics diagnostics from Chrome: hardware/software classification, renderer, backend, device/driver, feature and disabled-status details, and accelerated video capabilities. openclaw browser --json status returns the full structured payload. Passive status never launches Chrome just to collect these facts.
• stop closes the active control session and clears temporary emulation overrides. This applies even to attachOnly and remote CDP profiles, where OpenClaw did not launch the browser process itself. For local managed profiles, stop also stops the spawned browser process.
• start --headless applies only to that start request, and only when OpenClaw launches a local managed browser. It does not rewrite browser.headless or profile config, and is a no-op for an already-running browser.
• Local managed profiles run headless by default on every platform. OPENCLAW_BROWSER_HEADLESS=0, browser.headless=false, or browser.profiles.<name>.headless=false explicitly requests a visible browser; Linux then needs DISPLAY or WAYLAND_DISPLAY. Externally owned browsers keep their current mode.

If the command is missing

If openclaw browser is an unknown command, check plugins.allow in ~/.openclaw/openclaw.json. When plugins.allow is present, list the bundled browser plugin explicitly unless the config already has a root browser block:

{
  plugins: {
    allow: ["telegram", "browser"],
  },
}

An explicit root browser block (for example browser.enabled=true or browser.profiles.<name>) also activates the bundled browser plugin under a restrictive plugin allowlist.

Related: Browser tool

Profiles

Profiles are named browser routing configs:

• openclaw (default): launches or attaches to a dedicated OpenClaw-managed Chrome instance (isolated user data dir).
• user: controls your existing signed-in Chrome session via Chrome DevTools MCP.
• custom CDP profiles: point at a local or remote CDP endpoint.

openclaw browser profiles
openclaw browser system-profiles
openclaw browser system-profiles --browser brave
openclaw browser import-profile --browser chrome --system Default --into imported
openclaw browser import-profile --system "Profile 1" --into work --domains google.com,youtube.com
openclaw browser create-profile --name work --color "#FF5A36"
openclaw browser create-profile --name chrome-live --driver existing-session
openclaw browser create-profile --name remote --cdp-url https://browser-host.example.com
openclaw browser delete-profile --name work

Use a specific profile with --browser-profile <name> on any subcommand, for example openclaw browser --browser-profile work tabs.

On macOS, system-profiles lists real Chrome, Brave, Edge, or Chromium profiles available on the host. import-profile decrypts their cookies after one macOS Keychain/Touch ID consent prompt and injects them into a fresh OpenClaw-managed profile. It imports cookies only. Local storage and IndexedDB are unchanged. Some Google sessions use device-bound session credentials (DBSC) and can still require re-authentication after import.

Newer macOS versions can also ask whether the Gateway's Node process may access the browser's data. Until someone answers that prompt on the Gateway host, system-profile discovery and import time out while the rest of the Gateway keeps running.

When the macOS app uses a local Gateway, it can offer this import once and make the isolated imported profile the default for agent browsing. Import always requires an explicit click. Successful import or dismissal suppresses later automatic prompts, and Settings → General → Browser login remains available for re-import.

System-profile import is enabled by default. Set browser.allowSystemProfileImport=false to disable both CLI and agent-triggered imports. Import is host-local and cannot run through the browser node proxy.

Cookie sync to a remote Gateway

import-profile targets a managed profile on the same host. When your OpenClaw Gateway and agent browser run on a separate computer, use cookie-sync instead. It decrypts cookies on this Mac and pushes them into a managed profile on that remote Gateway over the operator connection:

openclaw browser cookie-sync --domains github.com,news.ycombinator.com --into work
openclaw browser --url wss://gateway.example.com cookie-sync --domains github.com --into work --watch

• --domains is required. Cookie sync copies live session cookies, so it never sends an unrestricted cookie jar. A missing or empty allowlist is a hard error.
• --into selects the target managed profile on the Gateway (default imported). --gateway/--url selects a remote Gateway (default is the configured or local one).
• --watch keeps the command running and re-pushes when the source Cookies database changes. The macOS Keychain secret is read once per watch session, so you approve a single consent prompt rather than one per change.
• Decryption is host-local (macOS only) and reuses the same allowlist and Keychain path as import-profile. Cookies are decrypted on this Mac and shipped over the existing TLS-pinned Gateway connection. No cookie values are printed.
• Some Google sessions use device-bound session credentials (DBSC) that stay tied to this Mac and can still require re-authentication after sync. For those sites, prefer driving the browser on the Mac itself through the browser node proxy.

The macOS app exposes the same capability under Dashboard → Settings → This Mac → Browser: an off-by-default toggle, an editable domain allowlist, and a target-profile field. When enabled in remote mode it supervises cookie-sync --watch for you against the connected Gateway and shows a live status row.

Chrome extension relay

openclaw browser extension path
openclaw browser extension setup --action inspect --json
openclaw browser extension setup --action install --json
openclaw browser extension setup --action verify --browser-profile chrome --json
openclaw browser extension install
openclaw browser extension install --no-store
openclaw browser extension install --json --wait-ms 60000
openclaw browser extension status
openclaw browser extension status --json
openclaw browser extension uninstall-host
openclaw browser extension uninstall-store
openclaw browser extension pair
openclaw browser extension pair --gateway-url wss://gateway.example.com
openclaw browser extension cdp
openclaw browser extension cdp --json

• extension setup is the shared host-local controller for CLI, TUI, and native desktop adapters. inspect is read-only installation discovery, install prepares native bootstrap, and verify authenticates the selected local relay. Its redacted JSON separates preparation, Chrome approval, and connection. Valid pending/blocked states exit 0; execution failures exit nonzero. It never treats a remote dashboard or SSH loopback URL as proof of a local browser host.
• extension install pre-registers the origin-locked native bootstrap host. Google Chrome on macOS can be prepared before its first launch; other supported browsers need an existing user-data directory. On macOS, setup then requests the official Store installation in Google Chrome for all profiles in that directory. Chrome discovers this at startup. Fully quit and reopen Chrome when convenient, then approve or enable OpenClaw. The command never restarts Chrome or bypasses approval. For other browsers and platforms, add OpenClaw from the Chrome Web Store. Linux supports automatic native pairing. Windows uses the self-contained OpenClaw.BrowserBootstrap.exe from the packaged CLI or Windows companion. A portable/local installer can pass --native-host-executable <absolute-path> to extension setup --action install or extension install. Setup checks owned context/ACLs and actual framed subprocess responses before user-level registry registration through the shared C# registration service. The CLI explicitly selects native Windows context; the Companion owns its managed-WSL mode. Conflicting contexts and unknown transport outcomes never trigger another writer or mode fallback. No executable or no proof means no automatic bootstrap; it never falls back to a script host or bypasses Chrome approval.
• extension install --no-store copies the stable development extension and registers the native host without creating a Store request. Existing requests are unchanged. Use the printed path for Load unpacked.
• extension status reports storeInstallRequests states (requested, missing, foreign, invalid) separately from storeDiscovered approval fields (enabled, awaitingApproval), approved unpacked IDs and paths, and native-host registration health. Local installation status does not prove a live relay connection. JSON output never includes a pairing string or relay key.
• extension uninstall-host removes only verified OpenClaw-owned native-host manifests and launchers. It does not remove the extension from Chrome. On Windows it delegates to the same registration service; --remove-store removes verified owned Store requests first. --native-host-executable and --browser-profile select the same explicit local context for status/removal.
• extension uninstall-store removes only OpenClaw-owned macOS Chrome Store requests (macOS only). Windows uses uninstall-host --remove-store rather than a separate Store writer. Chrome may remove an externally installed extension at its next startup. Native-host registration and the development copy remain intact.
• extension path is read-only. It prints the stable installed copy when present and the bundled source directory otherwise.
• extension pair remains the advanced manual flow. --gateway-url creates a direct remote-Gateway pairing URL. Non-loopback URLs must use wss://.
• extension pair --local-gateway --json lets desktop native helpers obtain the existing local pairing through the Gateway’s /browser/extension wake-up route. It requires a local Gateway configuration and cannot be combined with --gateway-url. The JSON contains a credential: consume it privately, never log it.
• extension cdp prints non-secret Browser Relay Authentication v2 metadata: the loopback browser/CDP endpoints, protocol version, key ID, and fixed challenge/complete binding. It never prints the relay key or an authorization header by default.

Automatic local bootstrap connects through the local Gateway's exact /browser/extension route so the first authenticated extension connection starts the lazy browser-control service. Keep openclaw gateway run or the managed Gateway service running. No separate browser request or prewarm is needed. Local OpenClaw and mcporter calls still use the profile relay port reported by extension pair or extension cdp after that wakeup. Browser-node pairings continue to use the relay on the browser-node host, while explicit --gateway-url pairings remain direct-remote and manual-only.

The advanced manual extension pair command without --gateway-url or --local-gateway retains the host-local /extension relay URL. With the native host installed, Automatic local setup enabled, and an extension build that supports relay wake-up, reconnecting can start a standalone relay on the saved pairing's configured port. This does not start Gateway browser control: authenticated CDP clients can use the standalone relay without a Gateway, but openclaw browser actions still require one. For source-checkout testing, load the managed unpacked copy from the same OpenClaw installation.

extension cdp --legacy-bearer is a temporary migration escape hatch. It prints the old Bearer header with a warning only while browser.extensionRelay.allowLegacyAuth=true. Otherwise it exits with an error without printing a credential. Use --json for machine output. Warnings remain on stderr so stdout stays valid JSON.

Setup, security model, and recovery steps: Chrome extension.

Run installation on the machine hosting Chrome. In the macOS app, Dashboard → Settings → This Mac → Browser → Set up Chrome on this Mac invokes the local CLI even when connected to a remote Gateway. The browser-based dashboard offers Store and documentation links instead.

If the extension already attempted automatic setup before the native host existed, Chromium retains that miss for the running browser process. Restart Chrome once, run extension install, then reopen the Store extension. Popup retries alone cannot recover that existing process.

Tabs

openclaw browser tabs
openclaw browser tab new --label docs
openclaw browser tab label t1 docs
openclaw browser tab select 2
openclaw browser tab close 2
openclaw browser open https://docs.openclaw.ai --label docs
openclaw browser focus docs
openclaw browser close t1

tabs returns suggestedTargetId first, then the stable tabId (such as t1), the optional label, and the raw targetId. Pass suggestedTargetId back into focus, close, snapshots, and actions. Assign a label with open --label, tab new --label, or tab label. Labels, tab ids, raw target ids, and unique target-id prefixes are all accepted. The request field is still named targetId for compatibility, but it accepts any of these tab references.

For HTTP CDP profiles, close confirms that the tab disappears before reporting success. If a page prompt keeps it open, resolve the prompt and retry. If the tab list cannot be read after the close request, confirm the tab's state in the browser before retrying; the close may already have succeeded.

Profiles configured with driver: "extension" can additionally return a numeric webExtensionTabId. It is scoped to the current browser runtime and is intended only for calls into Chrome's WebExtensions API. It can change after the browser or extension reconnects and is omitted for other drivers or when extension metadata is unavailable. Do not pass it to OpenClaw browser commands; keep using suggestedTargetId or tabId there.

Raw target ids are volatile diagnostic handles, not durable agent memory. Chromium can replace the underlying raw target during a navigation or form submit. OpenClaw then keeps the stable tabId or label attached to the replacement tab, when it can prove the match. Prefer suggestedTargetId.

Snapshot / screenshot / actions

Snapshot:

openclaw browser snapshot
openclaw browser snapshot --urls

Screenshot:

openclaw browser screenshot
openclaw browser screenshot --full-page
openclaw browser screenshot --ref e12
openclaw browser screenshot --labels

• --full-page is for page captures only. It cannot be combined with --ref or --element.
• existing-session / user profiles support page screenshots and --ref screenshots from snapshot output, but not CSS --element screenshots.
• --labels overlays current snapshot refs on the screenshot. On Playwright-backed profiles it works with --full-page (full-page overlay), --ref (element-clip overlay by ARIA ref), and --element (element-clip overlay by CSS selector). In element-clip modes labels are projected relative to the element. The response also includes an annotations array, omitted when empty. Each entry carries one ref's bounding box: ref, number, role, optional name, and box: {x, y, width, height}. Coordinates use the captured image's space (viewport, fullpage, or element-relative). existing-session profiles render a chrome-mcp overlay on page screenshots but do not use the Playwright projection helper and do not include annotations. CSS --element screenshots are unsupported there. Without Playwright or chrome-mcp, labeled screenshots are not available.
• snapshot --urls appends discovered link destinations to AI snapshots so agents can choose direct navigation targets instead of guessing from link text alone.

Navigate/click/type (ref-based UI automation):

openclaw browser navigate https://example.com
openclaw browser click <ref>
openclaw browser click-coords 120 340
openclaw browser type <ref> "hello"
openclaw browser press Enter
openclaw browser hover <ref>
openclaw browser scrollintoview <ref>
openclaw browser drag <startRef> <endRef>
openclaw browser select <ref> OptionA OptionB
openclaw browser fill --fields '[{"ref":"1","value":"Ada"}]'
openclaw browser wait --text "Done"
openclaw browser evaluate --fn '(el) => el.textContent' --ref <ref>
openclaw browser evaluate --fn 'const title = document.title; return title;'
openclaw browser evaluate --timeout-ms 30000 --fn 'async () => { await window.ready; return true; }'

press accepts named keys and shortcuts such as Escape, Control+Shift+T, and Control++. Common Esc, Return, Del, Ctrl, and Cmd aliases are normalized.

For managed browser profiles, select preserves option values exactly. Quote empty or whitespace-sensitive values, such as openclaw browser select <ref> "" or openclaw browser select <ref> " padded ".

evaluate --fn accepts a function source, an expression, or a statement body. Statement bodies are wrapped as async functions, so use return for the value you want back. Use --timeout-ms when the page-side function may need longer than the default evaluate timeout. browser.evaluateEnabled=false (default: true) disables both evaluate and wait --fn.

Action responses return the current raw targetId after action-triggered page replacement when OpenClaw can prove the replacement tab. Scripts should still store and pass suggestedTargetId/labels for long-lived workflows.

File + dialog helpers:

openclaw browser upload /tmp/openclaw/uploads/file.pdf --ref <ref>
openclaw browser upload media://inbound/file.pdf --ref <ref>
openclaw browser waitfordownload
openclaw browser download <ref> report.pdf
openclaw browser dialog --accept
openclaw browser dialog --dismiss --dialog-id d1

Managed Chrome profiles save ordinary click-triggered downloads into the OpenClaw downloads directory (/tmp/openclaw/downloads by default, or the configured temp root). Use waitfordownload or download when the agent needs to wait for a specific file and return its path. Those explicit waiters own the next download. Uploads accept files from the OpenClaw temp uploads root and OpenClaw-managed inbound media, including media://inbound/<id> and sandbox-relative media/inbound/<id> references. Nested media refs, traversal, and arbitrary local paths are rejected.

For remote browser nodes, OpenClaw stages private copies and normalizes filenames for portability, including Windows device names and trailing dots or spaces. File bytes remain unchanged.

Explicit download paths and trace stop --out preserve filename whitespace where the filesystem supports it. Quote these paths, for example openclaw browser download <ref> " report.pdf ". Output paths must stay inside their managed download or trace directory.

If saving a download fails, OpenClaw requests cancellation of the transfer and reports the original save error. Correct the output path or filesystem problem before starting a new download.

When an action opens a modal dialog, text output reports the block and pending dialog IDs. The JSON response returns blockedByDialog with browserState.dialogs.pending. Pass --dialog-id to answer it directly. Dialogs handled outside OpenClaw appear under browserState.dialogs.recent.

Batch actions:

openclaw browser batch --actions '[{"kind":"wait","timeMs":500},{"kind":"click","ref":"12"},{"kind":"type","ref":"23","text":"hello"}]'
openclaw browser batch --actions-file plan.json
openclaw browser batch --actions-file - --continue

openclaw browser batch sends a kind="batch" /act request with nested BrowserActRequest actions (wait, click, type, evaluate, ...) — not open/navigate/snapshot/screenshot, which are CLI subcommands, not /act kinds. --continue sets stopOnError=false (default stops on first error). --target-id scopes the whole batch to one tab. A failed nested action makes the command exit nonzero. Use --json to retain the ordered results response. See Browser batch CLI for the full contract (ref lifecycle, target id conflicts, error summary). batch is not supported on profile="user" / existing-session profiles.

If navigation or a closed page stops the batch, text output reports the action number and skipped count. Take a fresh snapshot before continuing with dependent actions.

--actions-file and --actions-file - stdin input are capped at 1,000,000 bytes. Split larger plans into multiple openclaw browser batch commands.

fill --fields-file, batch --actions-file, and batch stdin require valid UTF-8. Malformed bytes are rejected before any browser action runs. Save the input as UTF-8 and retry; valid Unicode values, including emoji and CJK text, are preserved.

fill --fields-file requires a regular file or a symlink to one. Pipes and other special files are rejected before dispatch. Fields files have no size cap.

State and storage

Viewport + emulation:

openclaw browser resize 1280 720
openclaw browser set viewport 1280 720
openclaw browser set offline on
openclaw browser set media dark
openclaw browser set timezone Europe/London
openclaw browser set locale en-GB
openclaw browser set geo 51.5074 -0.1278 --accuracy 25
openclaw browser set device "iPhone 14"
openclaw browser set headers '{"x-test":"1"}'
openclaw browser set credentials myuser mypass

Cookies + storage:

openclaw browser cookies
openclaw browser cookies set session abc123 --url https://example.com
openclaw browser cookies clear
openclaw browser storage local get
openclaw browser storage local set token abc123
openclaw browser storage session clear

Storage keys preserve surrounding whitespace. Quote the key in shell commands, for example openclaw browser storage local get " account ". Setting that key does not overwrite the separate account entry. Empty or whitespace-only keys remain invalid for set; omitting the key in get lists all entries.

Debugging

openclaw browser console --level error
openclaw browser pdf
openclaw browser responsebody "**/api"
openclaw browser highlight <ref>
openclaw browser errors --clear
openclaw browser requests --filter api
openclaw browser trace start
openclaw browser trace stop --out trace.zip

responsebody writes the bounded response prefix to stdout and warns on stderr when it is truncated. --json includes the response metadata and truncated flag without a separate warning. Use --max-chars to select the prefix limit.

Existing Chrome via MCP

Use the built-in user profile, or create your own existing-session profile:

openclaw browser --browser-profile user tabs
openclaw browser create-profile --name chrome-live --driver existing-session
openclaw browser create-profile --name brave-live --driver existing-session --user-data-dir "~/Library/Application Support/BraveSoftware/Brave-Browser"
openclaw browser create-profile --name chrome-port --driver existing-session --cdp-url http://127.0.0.1:9222
openclaw browser --browser-profile chrome-live tabs

The default existing-session path is host-only Chrome MCP auto-connect. If the browser is already running with a DevTools endpoint, pass --cdp-url so Chrome MCP attaches to that endpoint instead. For Docker, Browserless, or other remote setups where Chrome MCP semantics are not needed, use a CDP profile instead.

Current existing-session limits:

• Snapshot-driven actions use refs, not CSS selectors.
• Supported act requests use a built-in 60000 ms default when callers omit timeoutMs. Accepted per-call overrides set that action budget.
• click is left-click only.
• type does not support slowly=true.
• press does not support delayMs.
• hover, scrollintoview, drag, select, and fill reject per-call timeout overrides. evaluate accepts --timeout-ms.
• select supports one value only.
• wait --load networkidle is not supported (works on managed and raw/remote CDP profiles).
• File uploads require --ref / --input-ref and do not support CSS --element. Pass multiple paths when the page's file input accepts multiple files.
• Dialog hooks do not support --timeout.
• Screenshots support page captures and --ref, but not CSS --element.
• responsebody, download interception, PDF export, and batch actions still require a managed browser or raw CDP profile.

Existing-session action steps share one execution budget: filling and submitting with type do not each receive a fresh timeout. A conditional wait allows its explicit timeMs delay plus the action budget (with a 250 ms minimum) to satisfy the condition. A pure timer wait reserves the larger of timeMs and the action budget.

Navigation verification has a separate shared allowance of the action budget plus 1250 ms for scheduled delays. resize and close skip verification. Browser and tab preparation, execution, and final URL lookup share the overall request deadline. Internal calls and navigation checks do not renew it.

Remote browser control (node host proxy)

If the Gateway runs on a different machine than the browser, run a node host on the machine that has Chrome/Brave/Edge/Chromium. The Gateway proxies browser actions to that node. No separate browser control server is required.

Automatic routing prefers the Gateway host's browser and uses a single connected browser node only when local browser capability is unavailable. Use gateway.nodes.browser.mode to control this fallback and gateway.nodes.browser.node to explicitly select a node, including when the host has a browser. A stopped local managed browser with an installed executable still stays local.

Security + remote setup: Browser tool, Remote access, Tailscale, Security

Related

• CLI reference
• Browser