Unlike most MCP servers that wrap a single SaaS API (e.g. Gmail, GitHub), the Web MCP wraps the entire internet. It handles JS-heavy sites, auto-solves CAPTCHAs, and returns clean Markdown your model can use.
Free tier includes:
search_engine → search results from Google/Bing/Yandex
scrape_as_markdown → fetch any URL as clean, LLM-friendly Markdown (with CAPTCHA handling)
Quick start: https://docs.brightdata.com/mcp-server/quickstart/remote
I also wrote a blog post with the full background, challenges, and architecture:
Solving Web Access for Agents (For Free!):https://brightdata.com/blog/ai/web-mcp-free-tier
Would love feedback - what would you want to use live web access for in your agents?
meirk-brd•1h ago
Happy to answer anything here: design decisions, implementation details, or roadmap ideas. Really curious how the HN community would apply “internet access for agents” in real workflows.