A curated list of tools that empower AI agents to interact with the web — from cloud browsers and authentication to automation frameworks, enterprise packages, benchmarks, and search.
Cloud-hosted and open-source browsers designed for AI agents.
- Anchor Browser — Cloud browser built for AI agents. Handles sessions, CAPTCHAs, MFA, and proxies automatically. Free tier available.
- Steel — Open-source headless browser API for AI agents. Self-hostable or cloud. 6k+ GitHub stars. (GitHub)
- Browserbase — Cloud browser platform with session replay, stealth mode, and Playwright/Puppeteer support. Paid.
- Browserless — Scalable headless Chrome in the cloud. Paid.
- APIFY — Browser automation + web scraping platform. Paid.
Model Context Protocol servers that give LLMs (Claude, GPT-4, etc.) direct browser control.
- Playwright MCP — Official Playwright MCP server. Uses accessibility snapshots (not screenshots) for fast, reliable LLM-driven automation. Open-source.
- Steel MCP Server — Puppeteer-based MCP server backed by Steel cloud or local Docker. Open-source.
- Anchor Browser MCP — MCP server for Anchor Browser cloud sessions. Handles auth, CAPTCHAs, and concurrent sessions.
Tools that enable AI agents to access authenticated websites and sessions.
- Anchor Browser — Built-in MFA handling, cookie/session persistence, and fingerprint management for agents.
Autonomous AI agents that navigate and complete tasks on the web.
- Please (formerly MultiOn) — Consumer AI agent for web tasks (travel booking, forms, shopping). Rebranded from MultiOn in 2025.
- Jace.AI — AI agent for complex multi-step web workflows. Paid.
Tools and SDKs that power web automation with AI.
- Stagehand by Browserbase — AI-native browser automation SDK. Built on Playwright with
act(),extract(),observe(), andagent()primitives. Works with Claude, GPT-4, Gemini. (GitHub) - Browser-use — Python library connecting LLMs to browser actions. High-level task interface over Playwright.
- Skyvern-AI — Visual LLM-driven browser agent. Understands pages through screenshots rather than DOM.
- Agent-E — Hierarchical AI agent for browser automation tasks.
- LaVague — Natural language browser control using text language models.
- Tarsier by Reworkd — Vision utilities for web agents (element tagging, OCR).
- Notte — Framework for building web agents with planning, vision, and LLM reasoning.
- Axiom.ai — No-code browser automation bots. Paid.
- Browse.ai — Extract and monitor web data without code. Paid.
- ZeroStep — AI-powered test automation using plain English assertions. Paid.
Pre-built, open-source browser automation for enterprise platforms. Powered by Puppeteer and Anchor Browser cloud. Published as @browser-automation-hub/* npm packages.
| System | npm Package | Category |
|---|---|---|
| Epic EHR | @browser-automation-hub/epic-ehr-browser-automation |
Healthcare |
| Cerner Powerchart | @browser-automation-hub/cerner-powerchart-browser-automation |
Healthcare |
| Workday HCM | @browser-automation-hub/workday-hcm-browser-automation |
HR |
| ADP Workforce Now | @browser-automation-hub/adp-workforce-now-browser-automation |
HR |
| Salesforce Classic | @browser-automation-hub/salesforce-classic-browser-automation |
CRM |
| ServiceNow | @browser-automation-hub/servicenow-browser-automation |
ITSM |
| SAP Concur | @browser-automation-hub/sap-concur-browser-automation |
ERP |
| SAP Fiori | @browser-automation-hub/sap-fiori-browser-automation |
ERP |
| Oracle EBS | @browser-automation-hub/oracle-ebs-browser-automation |
ERP |
| PeopleSoft | @browser-automation-hub/peoplesoft-browser-automation |
ERP |
| Veeva Vault | @browser-automation-hub/veeva-vault-browser-automation |
Life Sciences |
| Veeva CRM | @browser-automation-hub/veeva-crm-browser-automation |
Life Sciences |
| Yardi Voyager | @browser-automation-hub/yardi-voyager-browser-automation |
Real Estate |
| Procore | @browser-automation-hub/procore-browser-automation |
Construction |
| SuccessFactors | @browser-automation-hub/successfactors-browser-automation |
HR |
→ View all 30 packages on GitHub · Postman Collections
Datasets and leaderboards for evaluating web agents.
- WebArena — Realistic web tasks across multiple sites. Open-source.
- WebArena Leaderboard — Live rankings.
- VisualWebArena — Multimodal extension of WebArena with image-based tasks. Open-source.
- WebCanvas — Online web agent evaluation with live environments. Open-source.
- WorkArena — Benchmark for enterprise web tasks (ServiceNow). Open-source.
- Mind2Web — Cross-task, cross-website generalization benchmark. Open-source.
- Bananalyzer by Reworkd — Web agent evaluation framework. Open-source.
Computer vision and multimodal approaches to web interaction.
- Skyvern-AI — Visual LLM agent that reads pages as screenshots. Open-source.
- VimGPT — GPT-4V driving a browser via keyboard shortcuts. Open-source.
Intelligent crawlers for data extraction at scale.
- FireCrawl — LLM-ready web crawling and scraping with Markdown output. Paid.
- Crawlee — Open-source web scraping and crawling library by APIFY. Supports Playwright, Puppeteer, Cheerio. Open-source.
- Spider — Fast AI web crawler with agent mode. Paid.
- Reworkd — AI-driven data extraction pipelines. Paid.
Search APIs for giving agents real-time web access.
- Tavily — Search API built for AI agents. Returns clean, LLM-optimized results. Paid (free tier).
- Exa — Neural search API for web content. Semantic search + keyword. Paid.
- Serper.dev — Google Search API for AI agents. Paid.
- Brave Search API — Independent search index with AI-ready API. Paid (free tier).
Tools for AI agents to handle CAPTCHAs and bot detection.
- Anchor Browser — Automatic CAPTCHA solving built into the cloud browser runtime.
- Capsolver — CAPTCHA solving API and browser extension. Paid.
- 2captcha — Human-powered CAPTCHA solving API. Paid.
- GraphQL.Chat — Explore and query GraphQL APIs with natural language. Free, closed source.
- GQLPT.dev — Generate GraphQL queries from plain text. Free, closed source.
Feel free to open a PR to add tools or update descriptions. This list is maintained by the community.