Browser CLI + Relay

Vibe Browser for OpenClaw

Use OpenClaw-style browser commands against your real logged-in browser session. vibebrowser-cli gives you a command-oriented surface for status, open, snapshot, click, type, and remote relay flows without switching to a disposable browser profile.

Real browser session

Reuse your actual tabs, cookies, logins, extensions, and saved state instead of spinning up a separate browser runtime.

Local or remote relay

Start locally on the same machine or point the CLI at a remote UUID when your browser is exposed through the Vibe relay.

Same package, two surfaces

The published npm package gives you both vibebrowser-cli for command flows and vibebrowser-mcp for MCP agents.

Command surfaces

Package the OpenClaw surface as explicit verbs

This is where a Tavily-style lesson actually applies. Not the branding, but the discipline: name the surfaces clearly so technical buyers understand what the route is for in under a minute.

Observe

Read browser state before acting

Command-oriented flows work when the operator can inspect the live page first. Status and indexed snapshots make the next action explicit instead of guessy.

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --json status
npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --json snapshot

Best for bug reproduction, portal triage, and any runbook that needs stable refs before the next click.

Act

Use explicit verbs against the live page

Open a page, click by ref, and type by ref with a thin command surface that still talks to the real browser session your team already uses.

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli open https://example.com
npx -y --package @vibebrowser/mcp@latest vibebrowser-cli click A12
npx -y --package @vibebrowser/mcp@latest vibebrowser-cli type A13 "hello world"

Good when you want shell-friendly control without turning the task into a full MCP integration yet.

Relay

Point the same flow at a remote browser

When the browser lives on another machine, keep the same command shape and swap in a UUID-backed remote relay instead of rewriting the workflow.

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --remote YOUR_UUID --json status

Useful for home-office browser access, remote operators, and hosted runners that still need the human browser state.

Escalate

Move to MCP without changing the package

The CLI is the thin command surface. When the job becomes richer, use the same published package to expose the browser to MCP agents.

npx -y --package @vibebrowser/mcp@latest vibebrowser-mcp browser --json status

That lets OpenClaw-style command flows and broader agent systems share one execution layer instead of fragmenting the product story.

Quick start for OpenClaw-style flows

These are the real published commands from @vibebrowser/mcp. Start with the CLI, then layer your OpenClaw skill or workflow on top.

Local commands

Check the local bridge

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --json status

Open a page

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli open https://example.com

Capture an indexed snapshot

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --json snapshot

Click by ref

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli click A12

Type into a field

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli type A13 "hello world"
Remote relay

If your Vibe extension is connected in remote mode, use the same CLI with a UUID:

npx -y --package @vibebrowser/mcp@latest vibebrowser-cli --remote YOUR_UUID --json status
Alias through the MCP binary

If you prefer one binary, the same command surface is also available as a subcommand on the MCP executable:

npx -y --package @vibebrowser/mcp@latest vibebrowser-mcp browser --json status

What this page is for

Use this route when you want a command-oriented browser interface, OpenClaw-compatible verbs, or relay-backed skill flows. If you need JSON MCP config blocks for Claude Code, Codex, Cursor, or VS Code, use Vibe Browser for Agents.

Where it wins

OpenClaw-style command flows for real operator work

This route should not read like a generic browser automation page. It is for operators who want a thin, explicit command layer on top of a real browser session.

Operator pattern

Browser bug reproduction loop

Open the failing route, capture a snapshot, click the suspect control, and hand the resulting evidence to a coding agent. This keeps the browser runbook explicit and replayable.

Deterministic command history
Snapshot refs an agent can reuse immediately
A cleaner handoff into richer MCP tooling when debugging expands

Operator pattern

Portal runbooks with a real session

For internal tools and authenticated portals, one human logs in once and the CLI reuses that exact browser state across repeated open, snapshot, click, and type steps.

No disposable browser profile to babysit
A shell-friendly interface for repeatable portal rituals
Lower ceremony than full custom automation stacks

Operator pattern

Remote browser, local commands

The browser can sit on a different machine while the command runner lives elsewhere. The remote relay keeps the flow thin without pretending the browser moved into the terminal.

Same CLI shape in local and remote modes
Better fit for distributed operators and hosted agents
A clear upgrade path into /mcp when more orchestration is needed

Command-oriented browser control, not a disposable browser

Vibe Browser for OpenClaw does not pretend to manage a fresh isolated browser lifecycle. It gives OpenClaw-style command flows access to the same real browser session your team already uses day to day.