DEV.co
MCP Servers · alibaba

page-agent

Page Agent is a TypeScript library that embeds an AI agent directly into web pages, enabling natural language control of web interfaces without requiring browser extensions, Python, or headless browsers. It works client-side, processes DOM text rather than screenshots, and supports most mainstream LLMs including locally deployed models.

Source: GitHub — github.com/alibaba/page-agent
25k
GitHub stars
2.1k
Forks
TypeScript
Primary language
MIT
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryalibaba/page-agent
Owneralibaba
Primary languageTypeScript
LicenseMIT — OSI-approved
Stars25k
Forks2.1k
Open issues50
Latest releasev1.11.0 (2026-07-03)
Last updated2026-07-07
Sourcehttps://github.com/alibaba/page-agent

What page-agent is

Page Agent is a JavaScript/TypeScript in-page GUI automation library that uses text-based DOM manipulation for LLM-driven web interactions. It operates entirely client-side, supports BYOL (bring your own LLM), and offers optional Chrome extension and MCP Server integrations for multi-page orchestration.

Quickstart

Get the page-agent source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/alibaba/page-agent.gitcd page-agent# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

SaaS AI Copilot Integration

Embed an AI assistant directly into SaaS products without backend rewrite. Ship natural language control in lines of code.

Enterprise Form & Workflow Automation

Automate repetitive multi-step workflows in ERP, CRM, and admin systems by converting 20-click processes into single natural language commands.

Accessibility & Voice Control

Make web applications accessible via natural language queries and voice commands, eliminating UI navigation barriers for users with accessibility needs.

Implementation considerations

  • LLM selection and API credentials must be managed; free demo API has testing terms and rate limits. Production use requires BYOL strategy.
  • Works only with text-based DOM; complex visual layouts or image-dependent interactions may require workarounds.
  • Client-side execution means agent reasoning runs in user's browser; consider token costs and latency for complex prompts.
  • Chrome extension is optional but required for multi-page workflows; deployment and permission management overhead applies.
  • No built-in authentication/rate-limiting layer for production SaaS copilots; add security controls upstream.

When to avoid it — and what to weigh

  • Server-side Automation Required — Page Agent is client-side only. If you need server-side headless browser automation or RPA across uncontrolled environments, use alternatives like Playwright or Selenium.
  • Multi-page Cross-domain Tasks Without Extension — Single-page scope without the optional Chrome extension. Workflows requiring coordination across separate domains or tabs need the extension deployed.
  • Offline or Air-gapped Deployments — Requires external LLM API access or locally deployed LLM. No offline-only fallback documented.
  • High-frequency Real-time Automation — DOM-based agent reasoning may have latency constraints for sub-second or high-frequency interactive tasks.

License & commercial use

MIT License. Permissive OSI license allowing unrestricted use, modification, and redistribution for commercial and private purposes, with no warranty.

MIT License permits commercial use without restriction. However, ensure your LLM provider (e.g., Qwen, OpenAI) permits commercial use for your use case. Page Agent itself imposes no additional commercial restrictions.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityLow
DEV.co fitStrong
Assessment confidenceHigh
Security considerations

Client-side execution means all LLM interactions and DOM state occur in user's browser; protect API keys and consider XSS attack surface if embedding in untrusted environments. No official security audit or disclosure program mentioned. Chrome extension, if deployed, widens attack surface. Ensure LLM API endpoints and credentials are protected. Unknown security hardening against prompt injection or malicious DOM structures.

Alternatives to consider

Browser-use (Python)

Server-side browser automation with Playwright backend. Better for server-side RPA, multi-page workflows, and headless deployments. Page Agent is derived from this project but optimized for client-side SaaS copilots.

Playwright / Selenium

Mature headless browser automation for enterprise testing and RPA. Better for uncontrolled environments, cross-browser testing, and offline scenarios. Requires separate infrastructure.

Anthropic's Computer Use

Vision-based agent for screenshot-driven web interaction. Better for complex visual layouts or non-text UIs. Different tradeoffs: multimodal LLM costs vs. text-based simplicity.

Software development agency

Build on page-agent with DEV.co software developers

Page Agent makes it easy to embed natural language control into any web interface. Start with our free demo, then integrate with your preferred LLM. Get technical guidance and custom deployment support from Devco.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

page-agent FAQ

Can I use Page Agent in production without the free demo LLM API?
Yes. BYOL (bring your own LLM) is fully supported. Configure any mainstream LLM (Qwen, OpenAI, Anthropic, etc.) via baseURL and apiKey. The demo API is for technical evaluation only; production requires your own LLM provider account.
Does Page Agent work offline or without internet?
Not documented. LLM API calls require external connectivity. Locally deployed LLMs are supported if accessible on-network, but no offline-only mode is described.
Can Page Agent automate cross-domain or multi-tab workflows?
Single-page only by default. The optional Chrome extension (Beta) extends reach to multi-page scenarios, but requires extension deployment and maintenance.
What are the token cost implications?
Depends on your LLM provider. Page Agent sends text-based DOM payloads, not screenshots, reducing token overhead. However, long DOM trees or repeated reasoning loops will incur per-token charges; monitor usage with your provider.

Custom software development services

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If page-agent is part of your mcp servers roadmap, our team can implement, customize, migrate, and maintain it.

Ready to Add AI Copilot Capabilities to Your Web App?

Page Agent makes it easy to embed natural language control into any web interface. Start with our free demo, then integrate with your preferred LLM. Get technical guidance and custom deployment support from Devco.