Get This Tool
Vessel Browser
Pricing
- Model
- Free
Summary
Most AI browser automation breaks the moment an agent needs to stay logged in, resume mid-task, or ask you before doing something irreversible — and you find out after the damage is done. Vessel is a local Chromium runtime built to close that gap: visible, stoppable, and designed for agents that need your sign-off on the risky steps.
Vessel exposes 80+ browser tools over MCP — navigate, click, fill, extract, diff, checkpoint, undo — so any MCP-compatible agent harness can drive a real browser without brittle Playwright scripts or headless guesswork. The visible-by-default design means you can watch, pause, approve, or roll back before something ships. Session persistence and named checkpoints mean the agent picks up where it left off instead of re-authenticating from scratch every run. The wall appears on complex, multi-site workflows that branch based on what the last page returned — the built-in skill system handles linear recipes well, but conditional logic across domains requires more than reusable instructions. Saved sessions, credential vaults, and page-change history are paid-only features, so teams running fully autonomous monitoring pipelines hit that ceiling fast.
Bottom line: Vessel is the right call when you need a supervised, auditable AI browser that keeps sessions alive and lets you intercept before actions commit — but if your workflow branches conditionally across a dozen domains and needs to run unattended overnight, the free tier runs short and the architecture asks you to be present.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Visible-by-default browser with pause, approve, and deny controls on every agent action, so you catch credential submissions or form commits before they execute rather than after.
- Named checkpoints and session restore so the agent resumes a logged-in, mid-task state across runs — eliminating the re-authentication loop that breaks most headless automation after the first session expires.
- 80+ structured MCP browser tools covering navigation, extraction, diffing, and undo, which means any MCP-compatible agent harness can drive a real browser without you writing and maintaining custom Playwright wrappers.
- MIT-licensed and fully self-hosted, so session cookies, credentials, and page data never leave your machine — which matters the moment a workflow touches accounts you do not want proxied through a cloud service.
- Bring-your-own-model support via OpenAI-compatible endpoints, so swapping the underlying model when costs shift or a local model improves is a config change, not a migration.
Cons
Sign in to edit- Saved sessions, credential vaults, and page-change history are paid-only features — teams building unattended monitoring pipelines on the free tier hit this wall the first time an agent needs to resume a session it did not start in the current run, and the options are upgrade or build session persistence themselves.
- The reusable skills system handles linear, site-specific routines well but has no native conditional branching — when an agent needs to take different paths based on what a page returned, teams add a separate orchestration layer on top of Vessel, which means two codebases to maintain instead of one; at that complexity level, teams with existing Playwright or Puppeteer infrastructure tend to stay there rather than migrate.
- The project carries 60 GitHub stars and community reports of 10K+ NPM downloads — a small signal base, which means production edge cases surface without much prior documentation, troubleshooting threads, or community-vetted workarounds to reference.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Linux, Windows, Mac
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-06-19T00:24:13.994Z
Best For
Who it's for
- AI agents needing visible, controllable browser access
- Supervised web automation with approval gates
- Persistent session and checkpoint workflows
- MCP-compatible agent harnesses
- Developers building agentic tools with local control
What it does well
- Research and summarize websites by browsing pages and collecting information
- Fill forms and complete repetitive web tasks autonomously
- Maintain logged-in sessions across multiple agent runs
- Monitor pages for changes with automated alerts
- Teach reusable custom skills for site-specific workflows
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Vessel Browser free?
- Vessel Browser is a paid tool. No permanent free tier is offered.
- Is Vessel Browser open source?
- Yes. Vessel Browser is open source.
- Does Vessel Browser have an API?
- Yes. Vessel Browser exposes a developer API. See the official documentation at https://quantaintellect.com for details.
- Can I self-host Vessel Browser?
- Yes. Vessel Browser supports self-hosting on your own infrastructure.
- What platforms does Vessel Browser support?
- Vessel Browser is available on: Linux, Windows, Mac.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Vessel runs a local Chromium instance exposed as an MCP server, giving AI agents a structured interface to navigate pages, click buttons, fill forms, extract tables, snapshot diffs, and restore named checkpoints — all inside a browser window you can watch in real time. The agent receives a cleaned semantic view of each page — forms, buttons, overlays, meaningful text — rather than raw HTML, so it can act on what the page actually means rather than what it literally contains. You connect Vessel to any MCP-compatible agent harness — Claude-style orchestrators, Codex-style harnesses, or a built-in chat assistant — and bring your own model keys or point it at a local OpenAI-compatible endpoint.
The standout differentiator is the control layer. Vessel is visible by default: a sidebar shows every action the agent takes, and you can pause risky steps, approve or deny them before they execute, and roll back to a named checkpoint if something goes sideways. This is not a logging feature bolted on after the fact — approval gates and undo points are first-class primitives in the MCP tool surface, which means the agent itself knows to pause and ask rather than proceeding when a step is flagged.
Vessel fits best in supervised workflows where an agent handles the tedious browsing steps — form fills, research sweeps, repetitive table extractions — while a person stays in the loop for anything consequential. It is MIT-licensed and self-hosted, so credentials and session state stay on your machine. The free open-source core covers the core browser tool set; saved sessions, credential vaults, vision features, and page-change history are paid-only features, meaning fully unattended monitoring pipelines or workflows that must survive across many days of agent runs require a paid tier.
Conditional branching across multiple sites — where what the agent does on step four depends on what it found on step two — is not well-served by the reusable skills system, which is designed for linear, repeatable flows. Teams that hit that ceiling typically layer a separate orchestration script on top, at which point they are maintaining Vessel plus their own branching logic as two distinct systems. Integration is via npm (`npm i @quanta-intellect/vessel-browser`) and the MCP protocol, with IPC handling the handoff between the agent layer and the Chromium runtime.