Problem
Teams often collaborate around the output of AI tools, but not the reasoning, constraints, and background that produced that output. That makes handoffs brittle and encourages people to paste giant walls of context back and forth.
AI · COLLABORATION · HACKATHON
ContextCollab
Shared context for teams working with AI agents.
ContextCollab was built during a Microsoft Hackathon at New York Tech Week 2026. The project came from a familiar team problem: AI agents are useful, but the context they build up often stays locked inside one person's chat.
The idea was to make that context easier to share without forcing everyone into the same agent or the same thread. Instead of copy-pasting long responses, teammates could pass along the useful parts of their agent context while keeping their own working memory, preferences, and flow intact.
Idea
ContextCollab treats agent context as something teams can share in smaller, intentional pieces: the goal, what has already been tried, decisions made, open questions, and useful artifacts. It keeps collaboration lightweight without flattening everyone's individual agent relationship.
Next
I'll add the full hackathon story, product flow, and more visuals here once I organize the rest of the project assets.