OpenKLAS MCP

An open-source FastAPI backend that wraps KLAS — Kwangwoon University's Learning Management System — and exposes it as a clean REST API with built-in MCP support for AI assistants.

Developed by @univerxe · Open to contributions

See what OpenKLAS does and how to set it up at openklas.com. To plug it into Claude.ai (or any MCP-compatible assistant) right now, add a custom connector with the URL https://mcp.openklas.com/mcp — you'll sign into KLAS once on the OpenKLAS login page and the full tool catalog becomes available.

---

Screenshots

Remaining lectures — One question. Every course. Aggregate progress. Claude scans every recorded-lecture board for you and reports back exactly which weeks remain in which subjects, with durations and deadlines included.	Homework overview — Every deadline, ranked by what actually matters. Ask once. Get pending, overdue, and submitted assignments across every subject in a single response, sorted by deadline, with overdue items flagged.
Assignment detail — Format, deadline, content checklist, all in one prompt. Stop opening four KLAS tabs to figure out what an assignment wants. Claude pulls the full task detail and surfaces it as a clean checklist.	Email drafting — One English sentence in. A polite Korean email out. Need to ask for a late-submission extension? Claude looks up the assignment, finds the right professor, and drafts a respectful Korean email ready to send.
PDF Q&A — Talk to your lecture PDFs like they're a tutor. Upload your slides once. Ask anything later. Voyage AI embeddings + pgvector + Ollama keep the answers grounded, per-user, isolated by design.	Lecture summarization — From an hour-long video to study-ready notes. Download → Whisper transcription → Claude-written summary, optionally saved straight to your Obsidian vault. Walk into class already prepped.

---

Tech Stack

Major Features

• MCP Server — every endpoint is a Claude tool via fastapi-mcp; connect to Claude.ai as a custom OAuth connector
• KLAS Auth — RSA-encrypted login, dual-token system (session token + JWT)
• Homework & Lectures — proxies assignments, lecture boards, course syllabi, team projects
• Recorded Lecture Pipeline — download → Whisper transcribe → Claude summarize → save to Obsidian
• RAG — ingest lecture PDFs, query them with Voyage AI embeddings + pgvector + Ollama
• Autocomplete — automated lecture progress completion

Contributing

Contributions are welcome! Feel free to open issues or pull requests.

1. Fork the repo
2. Create a branch (git checkout -b feat/your-feature)
3. Commit and push
4. Open a pull request