Open-source local browser agent Sonjari beats Playwright MCP in Claude Code benchmark
Stunning-Sherbet1853 · reddit · 2026-10-16
Sonjari is an Apache-2.0 local browser agent that works in your own Chrome via Ollama on your GPU, stops at logins/CAPTCHAs/payments, and centers on evidence: you can check which pages it actually read, numbers not found on read pages get flagged, and requirements are verified against outcomes. In a 27-row benchmark (find 3 products across 3 Korean shops, driven by Claude Code): Sonjari's browser tools scored 23/27 at $0.074/correct row; full hand-off 18/27; Claude Code alone 15/27 at $0.246; Claude Code + Playwright MCP 11/27 at $0.635. Only the Playwright gap is clearly significant (p≈0.002), and it's slow (6-7 min vs 1.5). Runs Swift-1.5-Qwen3.8-27B IQ3 128K on one RTX 3090; weaker on 8GB/6GB cards (37/63 and 10/63). Failures become tests (194 total); Windows + NVIDIA only; accepts tasks from Claude Code/Codex/Cursor over MCP.
More from coding & agent
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23
- How do you audit AI agent decisions in regulated workflows? A field survey — Ok_Way_1993 · 2026-10-23