Lumen: MCP for a learning platform
Lumen learning platform
The MCP server of the Lumen learning platform: course catalog, tutor questions with citations, a review queue and course drafts
- 93 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The server reads the platform database through a connection string and can create course drafts
- Ingesting by link calls YouTube, Notion and Google Docs
Install
- Works in
Manual install
claude mcp add lumen -- python -m app.mcp --transport stdioNeeds a running Lumen stack, with `LUMEN_MCP_AUTH_TOKEN` and `DATABASE_URL` in the environment.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads the platform database through a connection string and can create course drafts
- Ingesting by link calls YouTube, Notion and Google Docs
This is third-party code. Review the repository files before installing.
What it does
Lumen is a learning platform where the user runs the whole loop: states a goal in one sentence, an AI orchestrator builds a private 10-lesson course, and a course-scoped tutor answers with lesson citations. The repository shows a multi-agent system without frameworks such as LangChain, with public evals where weak scores are kept in the report. The platform's MCP server is described in `docs/mcp.md`. It offers nine tools: `list_courses` and `get_course` for the catalog, `search_lesson_content` for semantic search over lessons, `ask_tutor` for questions with citations, `list_my_due_reviews` and `grade_review_card` for the FSRS-6 review queue, `list_my_progress` for progress, `create_course_draft` for an AI course draft and `ingest_url_to_draft` for ingesting from YouTube, Notion and Google Docs. Rights split into public, user and instructor. There are two transports. Over stdio the client signs in with a static token in `LUMEN_MCP_AUTH_TOKEN`, issued by `make mcp-token`. Over streamable-http it uses OAuth 2.0 client-credentials. A running Lumen stack with PostgreSQL is required.
Who it is for. Developers and instructors who study a multi-agent learning system and want to connect its courses and tutor to Claude Desktop or Claude Code.
Good fit when
- You deployed Lumen and want to ask a course tutor straight from Claude
- You need the review queue and want to grade cards with FSRS-6
- You want to find a lesson fragment with semantic search
- An instructor wants a course draft from a Notion note or a YouTube video
- You study how an MCP server works on a self-built platform with OAuth
Not a fit when
- You want a ready cloud platform: the hosted demo was retired on October 30, 2026, so only a local run or your own deployment remains
- You do not want to run PostgreSQL and the repository stack: without them the server does not work
- You want a generic server for any course: the tools are tied to the Lumen database
Example request
Show my reviews due today, ask the course tutor how finite automata differ from regular expressions and show the lesson citations.MCP
- Transport
- stdio, http
- Authentication
- API key
Environment variables
LUMEN_MCP_AUTH_TOKENrequired, secret- Client token issued by make mcp-token
DATABASE_URLrequired, secret- PostgreSQL connection string of the platform
Limitations
This is the server of the Lumen project itself, with no standalone use: a running stack (`make up`, `make migrate`, `make seed`) and a PostgreSQL connection string are required. The hosted demo was retired on October 30, 2026; a walkthrough video, screenshots and redeploy instructions remain. The author calls the repository an example of working with agents at the protocol level, so it carries many architecture records and reports. Course build times (45 to 90 seconds for 10 lessons) refer to the earlier hosting on Groq. Instructor tools create course drafts, and ingesting by link reaches external services.
How to disable. Remove the `lumen` entry from the client configuration with `claude mcp remove lumen`. An issued client token can be removed in the database or reissued with `make mcp-token`.
FAQ
How many tools does the Lumen MCP server have?
Nine: `list_courses`, `get_course`, `search_lesson_content`, `ask_tutor`, `list_my_due_reviews`, `grade_review_card`, `list_my_progress`, `create_course_draft` and `ingest_url_to_draft`.
How do I connect Lumen to Claude Desktop?
Issue a token with `make mcp-token` and add a server to `claude_desktop_config.json` with the command `uvx`, arguments `--from lumen-backend python -m app.mcp --transport stdio` and the variables `LUMEN_MCP_AUTH_TOKEN` and `DATABASE_URL`.
What authentication is used?
A static token in `LUMEN_MCP_AUTH_TOKEN` over stdio, and OAuth 2.0 client-credentials over streamable-http with tokens issued via `POST /oauth/token`.
Does the hosted demo still run?
No, it was retired on October 30, 2026. A walkthrough video, screenshots and redeploy instructions remain, and the platform runs locally with the README commands.
What is FSRS-6 in the review tools?
A spaced-repetition algorithm. `grade_review_card` takes a rating from 1 (again) to 4 (easy), and `list_my_due_reviews` returns the queue of cards due.
Discussion
No questions or reviews yet.
Related
MCP serversShared company context for agents over MCP: an event graph with permissions, approvals and audit
A reference multi-agent student support assistant with an MCP server to a knowledge base and Bitrix24, over Telegram and MAX
One MCP gateway instead of dozens of servers: the agent finds a tool on demand and keeps no schemas in context
The mcpl CLI searches tools across several MCP servers and calls them from the terminal, with OAuth sign-in done by a human
MCP gateway of a self-hosted Reddit-style platform: agents publish posts with sources, vote and debate in the Arena.
MCP server of the self-hosted Synaplan AI platform: semantic search over your documents, long-term memory and chat
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Lumen: MCP for a learning platform (this card) | no | API key | 93 | |
| tutor-skills | no | 1,402 | tutor skills without a platform or course database | |
| Podcast to Course | no | 115 | creating a course from a podcast, with no tutor or review queue |
README in short
The README intro lays out the platform loop: goal, course build, learning with a tutor, publishing, cloning and bringing your own model key. It then covers the agentic layer: an orchestrator without frameworks, course-scoped RAG with citations, a log of agent steps and model calls, and eval suites with an LLM judge. It links to a walkthrough video, the architecture and the MCP server.