Hashbrown is a framework for building AI agents that run directly in the browser, purpose-built for Angular and React applications. It enables developers to bring multi-step task orchestration and external tool invocation into the frontend environment, allowing agents to directly manipulate browser UIs.
Key Features
- Browser-native agent execution with dedicated adapters for Angular and React, simplifying cooperation with components and routing
- Built-in and extensible connectors for models, retrieval, and external APIs to support common LLM workflows
- Permission boundaries and execution isolation for secure debugging and observability in the browser
- Modular architecture using adapters and workflow orchestration as core concepts
- TypeScript implementation with compatibility for modern frontend build chains and component systems
Use Cases
- Enhanced web assistants that coordinate interactive logic and model calls on the client side
- Automated form filling and data scraping agents with direct UI manipulation
- UI-focused task orchestrators that combine page interactions with LLM reasoning
- Browser-side entry point for RAG workflows where frontend-driven retrieval and generation are needed
Technical Highlights
- Developers can insert external models, caches, and retrieval components into browser-side execution paths while maintaining security boundaries
- Supports controlled coordination of model calls, page interactions, and external services on the client side
- Emphasizes compatibility with modern frontend tooling and component systems