Vessel lets an AI assistant open websites, click buttons, fill forms, remember sessions, and show every step in a real browser you can supervise.
Free open-source core. Premium adds saved sessions, vision, page-change history, credential vaults, and longer agent workflows.
$ npm i @quanta-intellect/vessel-browser
One prompt. Vessel searches Powell’s, chooses books, and adds them to a cart in a visible browser — so you can see exactly what the agent is doing.
Normal browsers are made for people clicking one step at a time. Vessel gives AI assistants a browser workspace where they can read, click, remember, and ask for your help when it matters.
Ask an AI to browse real pages, collect useful information, compare options, and bring back a clean summary instead of dumping tabs on you.
ResearchLet your assistant click through websites, fill fields, extract tables, and complete tedious browser steps while you keep the final say.
Web TasksSave browser sessions, tabs, cookies, highlights, bookmarks, and checkpoints so your AI does not have to start from scratch every time.
MemoryVessel is visible by default. See what the agent is doing, pause risky steps, approve sensitive actions, and roll back if something goes sideways.
ControlTrack important pages and let Vessel surface what changed, so your AI can help watch products, docs, dashboards, listings, or research sources.
MonitoringCreate reusable instructions for repeatable web tasks — research flows, shopping routines, site-specific steps, or anything your agent should remember how to do.
Custom SkillsUse Vessel with the built-in chat or connect it to an MCP-compatible agent harness. Bring your own model keys, local models, or OpenAI-compatible endpoint.
BYOK + MCPThe homepage stays plain: Vessel is a browser your AI can use while you supervise. Under the hood, it is a local Chromium runtime with MCP tools, persistent state, page intelligence, custom skills, checkpoints, and approval controls.
Vessel can run as a browser tool for an outside agent, or as a built-in chat assistant. Use OpenAI, Ollama, Mistral, Gemini, OpenRouter, llama.cpp, or any OpenAI-compatible endpoint.
External agent harness controls the browser via Model Context Protocol
Built-in sidebar assistant reads the page, uses browser tools, multi-turn history
Vessel still exposes the lower-level agent primitives when you need them: contextual tools, semantic hints, workflow state, macros, and progressive disclosure for complex browser automation.
On a login form, the agent only sees login-relevant tools
Elements annotated with workflow context and step progress
Workflow state maintained across page navigations
Login, search, checkout — common patterns as single actions
Tools revealed by layer — core, navigation, intelligence, diagnostic
Four ways to install. Pick whichever fits your workflow.
npm update -g to stay current.The free version is a real browser for AI-assisted web work. Premium adds the tools that make longer, messier, and more private workflows practical.
Best for trying Vessel, connecting an agent, and handling everyday browser tasks.
Install FreePlan options, trials, and billing are shown in app
Best for saved sessions, private accounts, visual pages, page monitoring, and longer agent workflows.
Install, Then Upgrade In App