# Vibe Mcp

> MCP server for browser automation - the only solution supporting multiple AI agents simultaneously (Claude Desktop, Cursor, VS Code, OpenCode)

- **Type:** MCP server
- **Install:** `agentstack add mcp-vibetechnologies-vibe-mcp`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [VibeTechnologies](https://agentstack.voostack.com/s/vibetechnologies)
- **Installs:** 0
- **Category:** [Web & Browser](https://agentstack.voostack.com/c/web-and-browser)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [VibeTechnologies](https://github.com/VibeTechnologies)
- **Source:** https://github.com/VibeTechnologies/vibe-mcp

## Install

```sh
agentstack add mcp-vibetechnologies-vibe-mcp
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Vibe MCP - Browser Automation for AI Agents

[](https://www.npmjs.com/package/@vibebrowser/mcp)
[](LICENSE)

MCP server for [Vibe AI Browser](https://vibebrowser.app) - the **only browser automation tool that supports multiple AI agents simultaneously**.

## Install prompt for AI agents (OpenClaw / Hermes)

Paste the block below into your **OpenClaw** or **Hermes** agent. Replace the last line with your remote value — a UUID or a `wss://relay.api.vibebrowser.app/` URL from the Vibe extension's **Settings → AI Agent Control → Remote (internet) → Relay access**. That's the only edit you make.

```text
You are setting yourself up to control my real Chrome through Vibe Browser. Work through the steps in order, and after each step confirm it worked before moving on. Do not claim a step is done until you have proven it.

1. Install the "vibebrowser" skill for your runtime:
   - OpenClaw:  npx -y skills add VibeTechnologies/vibe-mcp -s vibebrowser -a openclaw -g -y --copy --dangerously-accept-openclaw-risks
                then run `openclaw skills list` and confirm a "vibebrowser" row shows "ready" as installer registration.
   - Hermes:    npx -y skills add VibeTechnologies/vibe-mcp -s vibebrowser -y
                optional installer registration check: `npx -y skills list` shows a "vibebrowser" row.

2. Remember my remote so I only give it to you once. Save it to a file `vibebrowser-cli.md` in your persistent agent workspace root (the directory that holds your `memory/` folder — not the shell's working dir, not $HOME unless they are the same) as a line `remote: `; if your runtime can also write memory keys, additionally store it under `vibebrowser_remote`. On every later command — including the first message of a new conversation — load it back from that file and reuse it without asking me again.

3. Record a standing rule so you route correctly later: whenever I ask you to do anything in MY browser / MY Chrome / MY tabs / the page I'm on, use this vibebrowser skill (`npx -y @vibebrowser/cli@latest` with my saved remote) to drive my REAL browser — never your built-in, headless, or in-pod browser, never web-fetch instead, and never run `vibebrowser`. Save this rule wherever your standing instructions/memory live so it applies in future conversations.

4. Confirm you can control my Chrome, retrying on "Extension reconnecting" (wait 3s, up to 5x):
   npx -y @vibebrowser/cli@latest --remote "" --json status --wait-for-extension --wait-timeout 20000
   npx -y @vibebrowser/cli@latest --remote "" --json snapshot --format aria

Then report: (1) is extensionConnected true, (2) did the snapshot return my real page content?

My remote: 
```

## Why Vibe MCP?

| Feature | Vibe MCP | Playwright MCP | BrowserMCP |
|---------|----------|----------------|------------|
| **Multi-Agent Support** | Yes | No | No |
| Uses Your Browser Profile | Yes | No | No |
| Logged-In Sessions | Yes | No | No |
| No Separate Browser | Yes | No | No |
| Local & Private | Yes | Yes | Partial |
| Content Script Based | Yes | No | No |

### Multi-Agent Architecture

Vibe MCP is the **only solution that allows multiple AI agents to control the same browser simultaneously**. Run Claude Desktop, Cursor, VS Code Copilot, and OpenCode all at once - they all share control of your browser through our relay architecture.

```
Claude Desktop       Cursor          VS Code         OpenCode
     |                  |                |               |
     v                  v                v               v
 [vibebrowser-mcp] [vibebrowser-mcp] [vibebrowser-mcp] [vibebrowser-mcp]
     |                  |                |               |
     +------------------+----------------+---------------+
                        |
                        v
                  [Relay Daemon]  
Claude Desktop

Edit your Claude Desktop config file:
- **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
- **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`

```json
{
  "mcpServers": {
    "vibe": {
      "command": "npx",
      "args": ["-y", "@vibebrowser/mcp"]
    }
  }
}
```

Restart Claude Desktop after saving.

Cursor

1. Open Cursor Settings (Cmd/Ctrl + ,)
2. Go to "Features" -> "MCP Servers"
3. Click "Add Server" and add:

```json
{
  "vibe": {
    "command": "npx",
    "args": ["-y", "@vibebrowser/mcp"]
  }
}
```

Or edit `~/.cursor/mcp.json` directly.

VS Code (GitHub Copilot)

Add to your VS Code settings.json:

```json
{
  "github.copilot.chat.mcpServers": {
    "vibe": {
      "command": "npx",
      "args": ["-y", "@vibebrowser/mcp"]
    }
  }
}
```

Windsurf

Edit `~/.codeium/windsurf/mcp_config.json`:

```json
{
  "mcpServers": {
    "vibe": {
      "command": "npx",
      "args": ["-y", "@vibebrowser/mcp"]
    }
  }
}
```

OpenCode

Add to your `.opencode/config.json`:

```json
{
  "mcp": {
    "servers": {
      "vibe": {
        "command": "npx",
        "args": ["-y", "@vibebrowser/mcp"]
      }
    }
  }
}
```

Gemini CLI

Add to `~/.gemini/settings.json`:

```json
{
  "mcpServers": {
    "vibe": {
      "command": "npx",
      "args": ["-y", "@vibebrowser/mcp"]
    }
  }
}
```

OpenAI Codex CLI

Add to your Codex configuration:

```json
{
  "mcp": {
    "vibe": {
      "command": "npx",
      "args": ["-y", "@vibebrowser/mcp"]
    }
  }
}
```

### MCP command forms

Use the shortest direct package invocation for the MCP server:

```bash
npx -y @vibebrowser/mcp@latest --help
npx -y @vibebrowser/mcp@latest start --transport http
npx -y @vibebrowser/mcp@latest openclaw --remote "$VIBE_REMOTE_UUID"
```

Backward-compatible aliases still work when you need explicit binaries:

```bash
npx -y -p @vibebrowser/mcp@latest vibebrowser-mcp --help
npx -y -p @vibebrowser/mcp@latest vibe-mcp --help
```

### 3. Connect the Extension

1. Open Chrome with the Vibe extension installed
2. Click the Vibe extension icon in the toolbar
3. Go to Settings and enable "MCP External Control"
4. The status should show "Connected"

If the extension is not connected, `vibebrowser-mcp` can optionally fall back to
`chrome-devtools-mcp` (started in `--autoConnect` mode) when that package is installed.
This fallback runs once in the shared local relay daemon (multi-agent safe), so
both `vibebrowser-mcp` and `vibebrowser-cli` use the same backend instance.
When extension is connected, extension tools are authoritative.

### `--devtools` mode (drive your real Chrome over CDP)

Pass `--devtools` to either CLI to bypass relay/extension routing entirely and
drive your **real running Chrome** directly over the Chrome DevTools Protocol.
This backend (ported from the `chrome-use` skill) reads Chrome's
`DevToolsActivePort` file and autoConnects to your live profile — no extension,
no external MCP server, zero extra dependencies. Requires Chrome 144+; the
permission dialog fires once per profile.

`--devtools` exposes a focused tool set: `navigate`, `snapshot` (accessibility
tree with `@eN` refs), `click`, `fill`, `type`, `press_key`, `hover`, `scroll`,
`screenshot`, `eval`, `get_text`, `get_url`, `get_title`, and tab management
(`list_tabs`, `new_tab`, `select_tab`, `close_tab`). Use `snapshot` to get
`@eN` element refs, then pass them as selectors to `click`/`fill`/`type`.

Override the Chrome profile/channel with `VIBE_CHROME_USER_DATA_DIR` and
`VIBE_CHROME_CHANNEL` (`stable` | `canary` | `beta` | `dev`).

Not yet covered by `--devtools` v1 (addable later as more CDP domains are wired):
network request inspection, console logs, performance traces, Lighthouse, memory
snapshots, device emulation, dialog handling, file upload, and drag.

## Available Tools

| Tool | Description |
|------|-------------|
| `navigate_to_url` | Navigate to any URL |
| `go_back` / `go_forward` | Browser history navigation |
| `click` | Click elements on the page |
| `type` / `fill` | Enter text into inputs |
| `scroll` | Scroll the page |
| `take_screenshot` | Capture screenshots |
| `get_page_content` | Extract page text/HTML |
| `get_tabs` / `create_new_tab` / `switch_to_tab` / `close_tab` | Tab management |
| `keyboard_shortcut` | Press keyboard combinations |
| `web_search` | Search the web |

## How It Works

Default local mode (no flags):

```
Claude / Cursor / VS Code (stdio)
            │
            ▼
   [vibebrowser-mcp]
            │  ws://127.0.0.1:19888
            ▼
     Local Relay (auto-spawned)
            │  ws://127.0.0.1:19889
            ▼
     Vibe Extension (Chrome)
```

1. AI applications connect via MCP over stdio
2. `vibebrowser-mcp` connects to the local relay on port `19888`
3. The relay forwards commands to the extension on port `19889`
4. Results flow back to the agent

### Multi-Agent Mode

When multiple agents connect, Vibe MCP automatically spawns a relay daemon:

- First agent starts the relay (listens on ports 19888 and 19889)
- Additional agents connect to the relay as clients
- Relay multiplexes all agent requests to the single extension connection
- Each agent receives only its own responses

### Cloud OpenClaw -> Local Browser

If your agent runs in the cloud but you want it to control the user's real local browser, run `vibebrowser-mcp` in HTTP mode and connect it to the Vibe extension in remote relay mode. Pass either the extension UUID or the full WebSocket relay URL to `--remote`.

```bash
VIBE_REMOTE_UUID="YOUR-EXTENSION-UUID"
VIBE_REMOTE_URL="wss://relay.api.vibebrowser.app/YOUR-EXTENSION-UUID"

npx -y @vibebrowser/mcp@latest start --transport http --remote "$VIBE_REMOTE_UUID"
npx -y @vibebrowser/mcp@latest start --transport http --remote "$VIBE_REMOTE_URL"
```

This exposes a local MCP endpoint at `http://127.0.0.1:8788/mcp` by default.

When OpenClaw runs on a different machine (for example cloud-hosted), provide a reachable URL:

```bash
VIBE_REMOTE_UUID="YOUR-EXTENSION-UUID"
VIBE_REMOTE_URL="wss://relay.api.vibebrowser.app/YOUR-EXTENSION-UUID"
PUBLIC_MCP_URL="https://browser-bridge.example.com/mcp"

npx -y @vibebrowser/mcp@latest openclaw --remote "$VIBE_REMOTE_UUID" --public-url "$PUBLIC_MCP_URL"
npx -y @vibebrowser/mcp@latest openclaw --remote "$VIBE_REMOTE_URL" --public-url "$PUBLIC_MCP_URL"
```

You can print the exact OpenClaw-friendly setup with:

```bash
VIBE_REMOTE_UUID="YOUR-EXTENSION-UUID"
VIBE_REMOTE_URL="wss://relay.api.vibebrowser.app/YOUR-EXTENSION-UUID"

npx -y @vibebrowser/mcp@latest openclaw --remote "$VIBE_REMOTE_UUID"
npx -y @vibebrowser/mcp@latest openclaw --remote "$VIBE_REMOTE_URL"
```

Use `--remote ` with the default public relay, or `--remote ` when you need an explicit relay endpoint.

For direct browser CLI checks, always use `npx -y @vibebrowser/cli@latest`:

```bash
VIBE_REMOTE_UUID="YOUR-EXTENSION-UUID"
VIBE_REMOTE_URL="wss://relay.api.vibebrowser.app/YOUR-EXTENSION-UUID"

npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_UUID" --json status
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_URL" --json status
```

`--remote ` uses the default public relay. `--remote ` targets an explicit relay endpoint.

For the full walkthrough, see `docs/openclaw-local-browser.md`.

### OpenClaw-Compatible Browser CLI

`npx -y @vibebrowser/cli@latest` mirrors the OpenClaw browser CLI shape for the real local-browser path:

```bash
npx -y @vibebrowser/cli@latest sessions
VIBE_REMOTE_UUID="YOUR-EXTENSION-UUID"
VIBE_REMOTE_URL="wss://relay.api.vibebrowser.app/YOUR-EXTENSION-UUID"
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_UUID" status
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_URL" tabs
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_UUID" open https://example.com
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_URL" snapshot
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_UUID" click 12
npx -y @vibebrowser/cli@latest --remote "$VIBE_REMOTE_URL" type 23 "hello" --submit
npx -y @vibebrowser/cli@latest --devtools status
```

Use the MCP package for MCP server workflows and the CLI package for direct browser control:

- `vibebrowser-mcp` for MCP server, HTTP bridge, and helper commands
- `npx -y @vibebrowser/cli@latest` for OpenClaw-inspired browser control against the real Vibe-connected session

`@vibebrowser/cli` accepts the OpenClaw-style `--browser-profile` flag for compatibility and supports `--json` for machine-readable output. Unlike OpenClaw's managed `openclaw` browser profile, this CLI always targets the real Vibe-connected browser session.

Local-session selection:

- `npx -y @vibebrowser/cli@latest sessions` lists connected local browser sessions.
- `npx -y @vibebrowser/cli@latest --session  ...` targets a specific local session.
- If `--session` is omitted in local mode, the CLI uses the first connected session.
- In remote mode, pass either `--remote ` to use the default public relay or `--remote ` to use an explicit relay endpoint.

Snapshot behavior is tool-only (no legacy snapshot RPC shortcut):

- `snapshot` (default, `--format ai`) resolves via the `take_md_snapshot` tool — uses the content script's in-page markdown extractor. Fast and readable, but **may return empty for background tabs or complex SPAs** (Notion, Gmail) where the content script is unreachable or layout is not computed.
- `snapshot --format aria` resolves via the `take_a11y_snapshot` tool — uses Chrome DevTools Protocol `Accessibility.getFullAXTree` directly. **Reliable for all tabs including background tabs and SPAs.** Use this as a fallback when the default format returns empty or only a page title.

This keeps CLI behavior aligned with extension-supported tools and ensures page targeting works consistently with `--page-id`/`--pageId`.

For navigation-style operations, responses now include page content when page state changes:

- CLI `open` / `navigate` include `pageContent` in JSON output.
- MCP tool calls for navigation-style tools return text content that includes current page state (with snapshot fallback when needed).

### Install Prompt for AI Agents (OpenClaw / Hermes)

Use the single canonical copy-paste prompt at the top of this README:
**[Install prompt for AI agents (OpenClaw / Hermes)](#install-prompt-for-ai-agents-openclaw--hermes)**.
It installs the `vibebrowser` skill, saves your remote so you only give it once, and proves
control of your real Chrome.

For a deeper skill definition (when to prefer this over a managed/headless browser, the remote-recall rules, selector/snapshot guidance), install `openclaw/vibebrowser/SKILL.md` into the agent's skills directory.

### OpenClaw Integration

There are two ways to use Vibe with OpenClaw:

**Option A: Cloud OpenClaw controlling local browser**

If OpenClaw runs in the cloud but you want it to control your local browser:

1. Install the Vibe extension and enable **Remote** mode (see [docs/openclaw-local-browser.md](docs/openclaw-local-browser.md))
2. Start the local HTTP bridge: `vibebrowser-mcp openclaw --remote "$VIBE_REMOTE_UUID" [--public-url "$PUBLIC_MCP_URL"]` or `vibebrowser-mcp openclaw --remote "$VIBE_REMOTE_URL" [--public-url "$PUBLIC_MCP_URL"]`
3. Register the MCP URL in OpenClaw

**Option B: OpenClaw skill for local agents**

For OpenClaw agents that need your real browser context (logged-in sessions, existing tabs):

1. Copy the Vibe skill from this package to your OpenClaw skills folder
2. Use the extension UUID or full WebSocket relay URL with `--remote`
3. Use `npx -y @vibebrowser/cli@latest` commands in your agent prompts

The skill is located at [`openclaw/vibebrowser/SKILL.md`](openclaw/vibebrowser/SKILL.md) and provides:
- Full OpenClaw-compatible CLI commands (`status`, `tabs`, `snapshot`, `click`, `type`, etc.)
- Fallback-safe commands for DevTools-backed flows (`resize`, `upload`, `dialog`)
- `--json` output for machine parsing
- Environment-based configuration

See [docs/openclaw-local-browser.md](docs/openclaw-local-browser.md) for the complete walkthrough.

## Local LLM: `serve` Command

Run a local LLM with one command — no cloud API keys required. Automatically installs [Ollama](https://ollama.c

…

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [VibeTechnologies](https://github.com/VibeTechnologies)
- **Source:** [VibeTechnologies/vibe-mcp](https://github.com/VibeTechnologies/vibe-mcp)
- **License:** Apache-2.0

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-vibetechnologies-vibe-mcp
- Seller: https://agentstack.voostack.com/s/vibetechnologies
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.