Browser Agent Server

What it does

Provides a deployable full-stack browser automation system: a FastAPI backend that orchestrates Playwright/Xvfb-based browser sessions, a single-file dashboard with live screenshots and VNC pop-out, and multi-model LLM orchestration strategies (single, fallback_chain, planner_executor, consensus, council). The project bundles configuration and operational notes for running a browser-use server that can execute complex web tasks under LLM control.

When to use it

Use this when you need to automate browser interactions at scale or build an AI web agent that requires visual feedback and multi-model coordination — for tasks like automated testing, data extraction that needs a real browser, or agent workflows that benefit from council/consensus strategies to avoid loops and recover from failures. Also appropriate for sandboxed deployments where VNC/noVNC access is needed for debugging.

What's included

• Scripts: startup scripts and example agent_server.py, plus a dashboard HTML file (documented in the SKILL.md).
• References: operational notes and troubleshooting for Playwright/Chromium dependencies, Xvfb, and noVNC.
• Instructions: step-by-step setup covering system packages, venv creation, Playwright install steps, environment variables, and troubleshooting tips for browser launch failures and port/export conventions.

Compatible agents

Best for agents that can manage system-level deployment and run long-lived services (automation runners, devops-oriented agents, or assistants that can trigger shell provisioning). It assumes access to system packages and the ability to run Python services.