Log inSubmit server
Vessel Browser logo

Vessel Browser

Visible Chromium browser runtime for agent-led research, navigation, forms, and supervised web tasks.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Vessel Browser

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "vessel-browser": {
      "command": "npx",
      "args": [
        "-y",
        "@quanta-intellect/vessel-browser"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

We couldn’t confirm it starts correctly

We ran the install command but it didn't respond within our test window — this can mean a slow first-time install rather than a real problem.

npx -y @quanta-intellect/vessel-browser

No response to initialize.

This is an experimental automated check and can have false negatives — missing environment variables, a slow cold install, etc. It doesn’t necessarily mean something’s wrong. Last checked 5d ago.

Overview

Vessel Browser MCP server gives AI agents access to a visible Chromium browser for reading pages, following links, completing forms, and handling multi-step web tasks. It keeps agent activity in the browser window so a person can inspect actions, approve sensitive steps, redirect work, or take over manually. Persistent sessions, checkpoints, threads, bookmarks, and a shared run inbox help recover longer tasks. Reach for it when an agent needs to work on live websites rather than only answer questions about web content.

Use cases

  • Research information across multiple websites
  • Navigate pages and compare sources
  • Fill web forms with human approval
  • Resume interrupted multi-step browsing tasks
  • Inspect and undo agent browser actions

Key features

  • Visible Chromium browser for agent activity
  • Human approval and scoped action policies
  • Persistent sessions, threads, checkpoints, and run history
  • Structured bookmarks and inferred page schemas
  • Browser automation support for MCP clients
  • Undo for the latest successful mutating action

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 1d ago

Recent health check succeeded.

10/10 checks healthy over the last 49d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 12d ago via OSV.dev · @quanta-intellect/vessel-browser (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars137Stargazers on the source repository.
npm downloads998Package downloads in the last 30 days.
Last commit1mo agoMost recent push to the default branch.
Install checkInconclusiveDidn't respond in our test window — often a slow first install.
Directory activity3 viewsConfig copies, upvotes, and views on AllMCPs.

How Vessel Browser works

What Vessel Browser MCP server does

Vessel Browser MCP server provides an AI-operated web browser built on Chromium. Agents can inspect rendered pages, navigate between sites, extract page content, follow links, work with forms, and carry out longer browsing tasks. The browser remains visible while the agent works, giving a person the opportunity to review decisions and intervene.

The runtime is intended for tasks such as web research, comparing information across sources, monitoring page changes, and completing multi-step interactions. Its built-in assistant and external MCP or agent clients use the same browser foundation. The product also includes ordinary browser features such as tabs, bookmarks, reader mode, downloads, printing, and PDF saving.

How it works

Agent activity appears in a normal browser window instead of running as an invisible automation session. The Supervisor Sidebar provides a view of active work across up to ten tabs, including approvals, checkpoints, changes, research activity, conversations, and the durable Run Inbox.

Approval policies can apply once, to a run, or to a domain. A user can reject an action with steering, pause the work, or resume manual browsing. After a successful mutating agent action, the undo_last_action tool can restore the browser to the session snapshot captured immediately beforehand.

Longer work can be organized with named sessions, persistent conversation threads, checkpoints, notes, and structured bookmarks. Bookmarks may include intent, expected content, key fields, agent hints, and an inferred page schema. Vessel also infers page types, entities, form fields, and action-button intents for content extraction results.

Setup and configuration

Linux is the most mature installation target. A GitHub Releases AppImage can be downloaded, marked executable with chmod +x, and launched directly. The package can also be installed globally with npm install -g @quanta-intellect/vessel-browser and started with vessel-browser, or run without a global install using npx @quanta-intellect/vessel-browser.

A source installer is available through the repository's install script. Development from source uses Node.js 22, followed by npm install and npm run dev. After launching, open Settings with Ctrl+, to choose an AI provider and review advanced connection settings.

The built-in Chat Assistant supports Anthropic, OpenAI, Ollama, llama.cpp, Mistral, xAI, Google Gemini, OpenRouter, and OpenAI-compatible endpoints. Provider, model, and API key settings are configured in the application rather than through documented required environment variables.

Tools and capabilities

The Vessel Browser MCP server supports browser-oriented agent work, including:

  • Reading pages and extracting content
  • Navigating sites and following links
  • Filling forms and handling multi-step web tasks
  • Using persistent browser sessions, tabs, checkpoints, and threads
  • Saving and querying structured bookmarks
  • Reversing the most recent successful mutating action with undo_last_action
  • Supporting external agents and MCP clients
  • Inspecting console output, network requests, and MCP or agent activity through the Dev Tools Panel

The Research Desk can turn an approved plan into browser sub-agent work that collects source-backed claims and produces a markdown-exportable report. Research planning and execution have separate availability requirements in the product.

Limitations and notes

Linux has the most mature packaging support; macOS packaging is available from source. Vessel is described as active development software, so sensitive browsing and account access require care. Human supervision remains an explicit part of the workflow, particularly for approvals and actions that change external websites.

Some capabilities are not universally available: reusable Skills are marked Premium, and Research Desk plan approval, sub-agent execution, and report export require Vessel Premium.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Vessel Browser

Vessel Browser is an MCP server that connects AI agents to a visible Chromium browser and live web pages. Its main tools support page reading, navigation, link following, form interaction, multi-step browsing, and undoing the latest successful mutating action with `undo_last_action`, while people supervise activity in the browser.

Install it with npm using `npm install -g @quanta-intellect/vessel-browser`, then run `vessel-browser`; alternatively use `npx @quanta-intellect/vessel-browser`. Linux users can download the GitHub Releases AppImage, make it executable, and launch it. Source installation and development instructions are also provided.

Vessel Browser lets agents read pages, navigate websites, follow links, fill forms, compare information, and manage longer browsing tasks. Agents can also use persistent sessions, checkpoints, structured bookmarks, page schemas, and the `undo_last_action` capability.

Yes. Users can watch the agent in a visible browser, approve sensitive actions once or for a run or domain, reject and steer work, pause it, undo the latest successful mutating action, or take over manual browsing.

Vessel Browser supports Anthropic, OpenAI, Ollama, llama.cpp, Mistral, xAI, Google Gemini, OpenRouter, and OpenAI-compatible endpoints for its built-in Chat Assistant. Provider, model, and API key settings are configured in the application's Settings screen.

Related MCP servers

All in Browser Automation All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewVessel Browser AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/vessel-browser?style=directory)](https://allmcps.com/mcp/vessel-browser)
HTML Embed
<a href="https://allmcps.com/mcp/vessel-browser"><img src="https://allmcps.com/api/badge/vessel-browser?style=directory" alt="Vessel Browser on AllMCPs" /></a>