KCode

KCode is a platform for competing, learning and levelling up through weekly coding challenges. Users solve problems in an in-browser editor, submissions are graded against test cases, and results feed a progress dashboard and live leaderboard.

Practising for coding interviews and competitions is more motivating with regular challenges, instant feedback and people to compete against.

• Grade real code submissions against test cases.
• Track progress, accuracy and streaks over time.

• Weekly coding challenges by difficulty
• In-browser Monaco code editor
• Automatic grading for Python and JavaScript submissions
• Progress dashboard and live leaderboard
• Groups and discussions

ReactMonaco EditorTanStack QueryZustandTailwind CSSExpressSQLite

React SPA + Express REST API

React 19 SPA built with Vite, talking to an Express 5 REST API. Server state is handled by TanStack Query and client state by Zustand.

Data. SQLite via better-sqlite3, created and seeded automatically on first run.

API. REST endpoints for auth, challenges, code submissions, leaderboard and groups, plus an admin API for challenge CRUD, groups, user roles and submissions.

Auth. JWT auth with jsonwebtoken, passwords hashed with bcryptjs, and role-based access to the admin API.

Performance. One grader process handles every test case for a submission, rather than spawning a process per case.

Accessibility. Keyboard shortcut search and labelled navigation.

Security. Submissions run with a timeout; OS-level isolation (containers, seccomp) is the next step for running untrusted code at scale.

Trade-offs. Ran grading locally with the machine's own python3 and node rather than depending on an external sandbox service.

Building a test harness that converts inputs and outputs, including linked lists, between JSON and each language.

Grading untrusted code is as much a security problem as a correctness one.