Inspiration

We built Cursor Crew because we were tired of being blocked by team formation. We could not find teammates, so we wanted a way to execute ambitious ideas as solo builders and still compete at a high level in hackathons.

What it does

Cursor Crew lets one person coordinate multiple AI agents by voice, with each agent assigned a different responsibility like frontend, backend, infrastructure, and testing, so complex work happens in parallel instead of one prompt at a time.

How we built it

We designed and shipped this concept in about 2 hours by focusing on core demo moments:

  • 2×2 agent grid: four specialized agents visible at once
  • Live status and handoffs: who is speaking, who is working, and when work moves between agents
  • Dual history panes: command history and per-agent activity logs
  • Agent focus view: expandable panel with a Cursor-style TODO list and live preview of what that agent is building

Challenges we ran into

The biggest challenge was compressing a large multi-agent vision into a clear, interactive prototype under extreme time pressure. Working with a very small team footprint meant balancing product direction, UX decisions, and implementation speed at the same time.

Accomplishments that we're proud of

In just 2 hours, we turned a personal pain point into a working interactive demo with:

  • A clear user story (one person managing an AI crew)
  • Distinct agent responsibilities that feel like a real team
  • A believable orchestration flow that judges can understand quickly from the UI alone

What we learned

We learned that for multi-agent products, interface clarity matters as much as model capability. We also learned that speed comes from prioritizing the highest-impact interactions first, then polishing only what directly supports the story we want to tell in the demo.

What's next for Cursor Crew

Next, we plan to move from mock interactions to real execution:

  • Live voice input and transcription
  • Real agent task routing into Cursor via @cursor/sdk
  • Synchronized shared state across agents
  • Stronger conflict handling so Cursor Crew can function as a true AI-powered engineering team rather than just a prototype.

Built With

  • css
  • css3
  • cursor-agent-api
  • custom-orchestrator
  • docker
  • elevenlabs-speech-to-text-(scribe)
  • elevenlabs-text-to-speech-(turbo-v2.5)
  • es-modules
  • express.js
  • fetch-api
  • html5
  • icons
  • in-memory-state-store
  • jest
  • mock-agent-runtime
  • node.js
  • npm
  • react
  • regex-based-intent-routing
  • responsive
  • tabler
  • tsx
  • typescript
  • vanilla-javascript
Share this project:

Updates