Urgent.News

What's breaking now, across thousands of outlets.

Tech

Show HN: Turbo-Browser-MCP – Batch forms, DOM deltas and submit guards for MCP

Turbo-Browser-MCP is a Playwright-based Model Context Protocol (MCP) server designed to reduce LLM token waste and multi-turn latency in browser automation. It combines 25-33 fine-grained browser primitives into 10 batched tools, offering delta DOM snapshots, single-turn form prefill for React/Shadow DOM/iframe, direct disk file uploads, parallel tab sessions, network blocking presets, and a default human-in-the-loop Submit guard.

All benchmarks were conducted on headless Chromium locally using exact OpenAI o200k_base tokenization, with specific figures sourced from the benchmark.mjs file (npm run benchmark) and npm run benchmark. Eliminating 10 round-trips on a 10-field form results in a 7x reduction in browser execution time (from 574 ms to 82 ms) and a 10-fold decrease in separate LLM inference round-trips (timeframes ranging from 2-5 seconds per turn during agent loops) plus approximately 71,800 replayed history tokens.

Turbo-Browser-MCP charges an LLM round-trip for every snapshot, field fill, selection, upload, and verification check, in contrast to standard MCP servers. Each turn re-sends the expanding conversation history back into the model. The server features a prefilled form in a single MCP call, while simultaneously flushing microtasks and macrotasks between controlled inputs.

To run the benchmark suite, users must have Chromium and Playwright installed. The environment and methodology involve headless Chromium via playwright-core, with the exact OpenAI o200k_base tokens measured using Python tiktoken (approx-bpe fallback if unavailable). Cumulative workflow tokens model standard chat-history replay across N sequential turns, calculated as the sum from 1 to N of (N - i + 1) multiplied by Ti, plus a one-shot baseline where the first turn captures context before acting.

Turbo-Browser-MCP is compared against installed MCP JSON schemas on disk with compact parameter schemas lacking annotations. Turbo-Browser-MCP is intended for rapid multi-step UI automation, forms, and scraping with minimal token overhead. In situations where Chrome DevTools MCP is needed alongside Turbo-Browser-MCP, the Chrome DevTools MCP provides additional features such as Lighthouse audits, performance trace analysis, CSS computed style panels, V8 heap snapshots, and a built-in WebSocket in DevToolsActivePort for remote debugging in Chrome 136+/144+.

Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at github.com →

More in Tech

Beyond Black-Box Prompts: Building an Explainable Multi-Agent Financial Intelligence Platform with FastAPI & React

In quantitative finance and retail investing, treating Large Language Models as a single "oracle" prompt is dangerous. When you ask a single LLM whether a stock is a buy, it frequently exhibits…

  • Vibe-Finance platform combines multi-agent financial intelligence with explainability
  • System uses FastAPI backend, RAG over corporate filings, React/Vite/Tailwind client
  • Three specialized agents prevent single-point failure and bias in investment decisions

I Told My Agent Who Sent the Message. It Mostly Stopped Saying We Were Talking.

This is the line: AUTOMATED SCHEDULER CHECK: sender=scheduler, not Keniel; Keniel is not participating in this turn. I added it to a frozen copy of my agent's scheduled self-audit request.

  • Agent incorrectly claimed conversation while human not participating
  • Corrected sender line improved accuracy, p-value 0.0000002
  • Advice to check sender info in scheduled messages

Coalition Govt Reviews Its Future

President Asif Ali Zardari met Prime Minister Shehbaz Sharif at his residence in Model Town, Lahore, where the two leaders … Read More The post Coalition Govt Reviews Its Future appeared first on…

More from Sunday 15 November →