browser-use: Agent Prompt Structure

Live demo of a low‑cost LLM browser agent signing up for thousands of newsletters, scraping varied sites, and exposing its Langfuse prompt structure and patterns.

Python Gemini Vertex AI Langfuse Browser

Overview

I’ve spent $ 5000 on LLM tokens for browser-use, the agent for using web browsers.
Some tasks included signing up to 20k different newsletter forms (for an aggregated AI newsletter project) or gathering data (kind of like a general purpose, universal scraper).
I’ll show how the agent interacts with websites it hasn’t seen before.
I would also love to present how the agent prompt is structured by peeking inside Langfuse, and discuss prompting patterns with the community.

Links

https://gralio.ai
Gralio: AI-driven screen capture identifies process inefficiencies for automation.
https://github.com/browser-use/browser-use
Python library for AI agents automating browser tasks with Playwright.

Tech stack

Python

Python: The high-level, general-purpose language built for readability, powering everything from web backends to advanced machine learning models.

Python is the high-level, general-purpose language prioritizing clear, readable syntax (via significant indentation), ensuring rapid development for any team . Its ecosystem is massive: use it for robust web development with frameworks like Django and Flask, or leverage its power in data science with libraries such as Pandas and NumPy . The Python Package Index (PyPI) provides thousands of community-contributed modules, offering immediate solutions for tasks from network programming to GUI creation . The language is actively maintained by the Python Software Foundation (PSF), with the stable release currently at Python 3.14.0 (as of November 2025) .

https://python.org

View projects
Gemini Vertex AI

Vertex AI is the unified, fully-managed platform for building, customizing, and deploying Google's Gemini multimodal models and 200+ foundation models at enterprise scale.

Gemini on Vertex AI delivers Google's most capable multimodal models (e.g., Gemini 3 Pro, 2.5 Flash) directly into your enterprise workflow. Use Vertex AI Studio to rapidly prototype with text, images, and video, then leverage the Model Garden to access over 200 proprietary and open-source models. The platform provides a complete MLOps toolset—including Vertex AI Pipelines and Evaluation—to manage, govern, and scale your generative AI applications with Google Cloud's security and data integration (BigQuery) built-in.

https://cloud.google.com/vertex-ai

View projects
Langfuse

Langfuse is the open-source LLM engineering platform: gain full observability, manage prompt versions, and run production-grade evaluations.

Langfuse delivers the essential LLM engineering stack. It's an open-source platform for full-lifecycle management of your AI applications (agents, chains, etc.). Use its comprehensive tracing (OpenTelemetry-based) to debug complex, non-deterministic interactions and track exact cost/latency metrics. The system provides robust prompt management (versioning, A/B testing) and flexible evaluation tools to measure output quality and monitor production health. Integrations are native: connect with Langchain, OpenAI, and LlamaIndex via Python/JS SDKs for immediate control and clarity over your LLM deployment.

https://langfuse.com/

View projects
Browser

The client application that fetches, interprets, and renders content (HTML, CSS, JavaScript) from web servers for user display.

A browser is your primary interface to the World Wide Web, operating on a client-server model: it sends HTTP/HTTPS requests for files and uses a rendering engine to translate the returned data into a visual page. This process involves interpreting HTML structure, applying CSS styles, and executing JavaScript for dynamic functionality. Major players like Google Chrome, Mozilla Firefox, Apple Safari, and Microsoft Edge dominate the market, with Chrome currently holding a significant majority share. The core function remains consistent: transforming raw code into the interactive experience you see on your screen.

https://en.wikipedia.org/wiki/Web_browser

View projects

Related projects

MCP – Future of agents

Poland

The talk explains the MCP protocol, its operation, and demonstrates practical use cases, illustrating how it enables advanced…

telli's internal agents infrastructure

Berlin

Learn how we built an internal infrastructure providing all employees access to powerful coding and debugging agents for…

🕵🏻‍♂️ Secret Agent 🕵🏻‍♂️

Toronto

This talk covers secure secret management for crypto-native AI agents, enabling pay-as-you-go LLM access, key integration, and centralized…

Private Agents in your browser

Chicago

Demonstrates how to run small-language-model agents directly in the browser, enabling low-latency, cost-effective agentic tasks on a laptop…

Multi-Agent Systems Solving Complex Tasks in IT Environments

Poland

Explore a LangChain-based multi-agent architecture that integrates tools and graph knowledge retrieval to automate complex IT support and…

The Browser is Dead, Long Live AI Browser Agents

San Francisco

Examines core technical challenges of AI browser agents, comparing architectural designs, data handling, latency, security, and scalability tradeoffs…