PawBrowse: open-source MCP halves browser agent round-trips to one per action
Far-Round2092 · reddit · 2026-09-27
Most browser agents drive off screenshots, so every action takes a perceive-then-act pair of full model inferences, doubling tool calls, latency, and token costs. The open-source PawBrowse instead hands the agent a compact table of clickable elements (real names, values, state); refs are stable and every action returns a fresh table, so the agent acts in one round-trip.
- Measured (same task, same model — booking a 4-star Lisbon hotel on Booking.com): 5 agent round-trips vs 10, roughly 2x faster in a round-trip-bound loop and cheaper in tokens
- Drives your real, logged-in Chrome with no remote-debug port, no relaunch, no second model, no API key
- MIT-licensed, built with Claude Code, live on the Chrome Web Store + npm, fully reproducible from the repo
More from coding & agent
- Personal agents from big labs vs one shared company agent: which design wins? — AshishKulkarni1411 · 2026-11-16
- WebArena-Pro launches: 300-task multimodal web-agent benchmark across 20 apps — xhluca · 2026-11-16
- Self-improving agent papers this week: agents learn the benchmark, not the task — TheTuringPost · 2026-11-16
- 1,000 agents, one week, 630k lines: HF builds a weights-free Python chatbot — _lewtun · 2026-11-16
- One-person content agency targets $30K/mo with 11 AI agents — full playbook — VibeMarketer_ · 2026-11-16
- Hugging Face releases definitive fine-tuning guide for 3D generation model TRELLIS.2 — ariG23498 · 2026-11-16