← sources

Building a web layer for AI agents. If you use Firecrawl, Browserbase, TinyFish or similar, what's the one thing it must not get wrong?

source: reddit (reddit:AI_Agents)published 2026-11-14 14:49ingested 2026-11-15 02:33

retained text

I'm a solo dev building a layer that lets agents read pages (clean markdown), scrape, and take actions on sites, with a human handoff when the agent hits a CAPTCHA, OTP or login.



Before I go further, I want to learn from people who already run these tools in production.



If you use Firecrawl, context\_dev, TinyFish, Browserbase, Browser Use or your own setup:



1. What's the one feature you couldn't live without?

2. What made you pick your current tool over the others?

3. How do you actually call it day to day: REST API, SDK, MCP, CLI?

4. What still frustrates you about it, or what did you have to build yourself?

5. Do you stick to one tool, or do you combine several? Would you add another one to your stack if it did one thing clearly better?



I'm trying to figure out whether there's still a gap worth building for, from the point of view of people who use these tools every day. Honest answers help a lot.

document fetched 2026-11-15 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xw46ym/building_a_web_layer_for_ai_agents_if_you_use