Studybench

Studybench — an agent-native study dashboard with visible WebMCP tools and review progress.

Inspiration

Students often know what they should study but not what to do next. Studybench turns exam preparation into a focused, adaptive workspace where progress, review, planning, and agent actions share one visible state.

What it does

Studybench shows mastery by topic, cards due for review, exam countdown, and readiness at a glance. Students can start a review session, reveal answers, self-grade cards, and focus on weak topics such as Semaphores & Mutexes or Deadlock Handling.

The right-side agent workspace makes the product agent-native. A student can ask the agent to quiz them on their weakest topic, turn notes into cards, or rebuild a study plan. The Tools and Calls views expose the structured WebMCP actions so the student can see what the agent can do and keep control of the workflow.

How we built it

The project is a lightweight JavaScript web app deployed on Vercel. It uses a local-first study model for cards, topics, mastery, review sessions, plans, and activity. The agent surface is designed around WebMCP-style structured tools and observable calls rather than an opaque chat-only experience.

Challenges and learnings

The main challenge was making the agent useful without hiding the study state or taking control away from the student. We learned that a good agent-native interface needs explicit tools, visible calls, predictable state updates, and a manual path for every important action.

What's next

Future improvements include importing notes, richer spaced-repetition scheduling, collaborative study plans, and connecting the same structured tools to more browser agents.