Launched this week
DLBrowser
**Self-healing web access for AI agents**
16 followers
**Self-healing web access for AI agents**
16 followers
DLBrowser gives AI agents reliable web access in one line. A self-healing engine escalates through 5 backends to beat Cloudflare, captchas & JS-heavy pages — with per-run cost tracking and transparent credits. MCP-native, MIT, self-hostable. Works with Claude, Cursor, Windsurf & more.
Would love to see a simple dashboard for reviewing past runs showing which backend actually got used and why, so I can spot when the fallback chain is hitting a slower tier more often than it should.
finally something that handles captchas without me babysitting the agent. dropped it into my cursor workflow and the cost tracking was honestly the part that sold me.
finally tried it with a stubborn cloudflare page and it actually pushed through on the third backend without me babysitting it. the per-run cost line is genuinely useful for keeping my agent spend in check.
honestly the cost tracking per run is a nice touch, but it would be super useful to have a built-in retry budget or fallback summary so you know which backend actually succeeded and how many credits got burned along the way when things go sideways.
Finally something that doesn't choke on the first Cloudflare wall it hits, the multi-backend fallback actually worked on a stubborn site I'd been scraping manually for days.
The per-run cost tracking sounds super useful. One thing I'd love to see is a simple shared cache layer so if multiple agents hit the same page within a short window they don't each pay for a fresh fetch, would cut costs noticeably for teams running parallel agents on overlapping research.