Remember when you asked your AI to look something up on the web and it either returned a wall of unusable text, or just... didn't work? Yeah, we remember too. Not our finest hour. That's fixed now!

Two tools. Rebuilt from the ground up.

→ web_search – searches the web for any keyword and returns actual, useful results
→ web_scrape – reads any URL and extracts the content in clean, readable markdown

Simple as that. No more context overload. No more silent failures.

What changed under the hood:

We migrated both tools off our old infrastructure and rebuilt them properly. The old versions were held together with good intentions and a few too many workarounds. The new ones are lean, fast, and built to handle real workloads.

We also took Anthropic's own guidance on tool design seriously. That means your AI agents are now much better at knowing which tool to use and when, without you having to babysit the process.

The short version:

• Web research that actually works

• Clean content extraction from any URL

• No more oversized responses breaking your workflow

The update is live. Go try it.