Do AI agents need crawling, search, or screenshots first?

I’m building a web data API for AI agents and I’m trying to understand what builders actually need most.

The problem I keep running into is that “give an agent access to the web” quickly becomes several separate pieces:

- Fetching simple pages

- Rendering JavaScript-heavy pages

- Turning pages into clean Markdown

- Searching first before crawling deeper

- Capturing screenshots for verification

- Tracking API usage, logs, and costs

I’m currently working on AnyCrawler, which combines crawl, search, screenshots, API keys, logs, usage, and credits into one API/console.

But I’m more interested in the workflow question:

If you’re building agents, research tools, data products, or automations, which part is the biggest pain today?

1. Getting clean page content

2. Handling JS-rendered pages

3. Finding the right sources first

4. Visual verification with screenshots

5. Monitoring usage and cost

Curious what people here are doing in production.