Show HN: Browser Harness – Gives LLM freedom to complete any browser task
Browser Harness just removed the guardrails. Instead of predefined tools, it gives LLMs raw CDP access—and they're writing their own functions mid-task.
Why it matters
A fundamentally different approach to browser automation: instead of wrapping Chrome in constrained APIs, give LLMs direct access to the Chrome DevTools Protocol and let them solve edge cases autonomously. This shifts the paradigm from deterministic heuristics to emergent agent behavior.
The key facts
7 to knowOpen-source Browser Harness library removes framework constraints on LLM browser control
LLMs given raw CDP (Chrome DevTools Protocol) access with minimal helper abstractions
Agent self-corrected by writing upload_file() function using DOM.setFileInputFiles mid-task without explicit implementation
Demonstrates agent autonomy on complex tasks: Stockfish gameplay, Tetris world record, JavaScript drawing
Contrasts with existing approaches (Playwright MCP, Browser Use CLI, agent-browser) that use predefined function wrapping
Silent failure mode risk eliminated: LLM has full context of why tools fail vs. thinking click() succeeded when it didn't
Installation via Claude Code: one-line setup integration
Go to the source
Hacker Newsgithub.com
Publisher excerpt: Hey HN, We got tired of browser frameworks restricting the LLM, so we removed the framework and gave the LLM maximum freedom to do whatever it's trained on. We gave the harness the ability to self correct and add new tools if the LLM wants (is pre-trained on) that. Our Browser Use library is tens…