Skip to Main Content

Long running browser agents struggle when dealing with custom workflows in complex UIs. Watch and Learn is a multi-modal context engine which allows agents to learn from user and agent demonstrations. Each user query is converted to a VoyageAI embedding and queried against existing demonstrations from MongoDB. These demonstrations are then passed into the agent's context. The result is reliable, long-running agents.

Team

jamjam112Gii2000+1

Placement

Finalist

More from Agentic Orchestration and Collaboration Hackathon

Mongrate is a multi-agent coding system that migrates an app end-to-end from Postgres to MongoDB by having specialized agents analyze schema + code, design a…

Moongrate project preview

We built Polaris, an open source threat intelligence system that keeps agent workflows aligned with current vulnerabilities. It uses a static 2025 BERT…

www.loom.com/…

SOGA: Smart Orchestration Glasses Agent Voice-controlled AI agent orchestration through smart glasses. Spawn specialized agents (Scout, Analyzer…

drive.google.com/…

A multi-agent system that processes real-time video, builds 3D models to simulate the problem using depth estimation, and computes optimal paths for first…

www.loom.com/…

JustPrice is your AI Healthcare Advocate, designed to combat predatory medical pricing. Although federal law mandates transparency, hospitals often obscure…

www.youtube.com/…