MyCourseVille MCP Server
Enables an AI agent to access a user's MyCourseVille account to retrieve assignments with deadlines, course materials, released grades, and a feed of what has changed since the last sync.
README
MyCourseVille MCP Server
An MCP server that gives an AI agent access to your MyCourseVille account: new assignments with their deadlines and descriptions, course materials, and released grades — plus a "what changed since last time" feed.
Personal tool for your own account data. Keep the poll interval polite and never commit your credentials.
How it works
MCV has an official OAuth2 API (/api/v1/public/get/*), but it requires registering an
app to get a client_id/client_secret, and its endpoint names aren't publicly
documented. This server instead logs in exactly like the MCV web client does — through the
IT Chula SSO button by default, or the plain MCV form — then reads the same server-rendered
course pages the site serves to a browser: the course home for materials, /assignment for
deadlines, and /portfolio-<student id> for scores. See
docs/ENDPOINTS.md.
Fetched data lands in a local SQLite snapshot. Every sync diffs against that snapshot and
appends to an events log — that log is what whats_new reads, so the server can tell you
what appeared since you last looked. A background poller refreshes on an interval so a new
assignment is noticed even when no chat is open.
On "new": MCV doesn't reliably expose a posted-at timestamp, so first_seen — when
this server first observed an item — is the proxy. The first sync backfills everything
quietly rather than reporting your entire semester as new.
Setup
uv sync --extra dev
cp .env.example .env # then fill in MCV_USERNAME / MCV_PASSWORD
uv run mcv-mcp --selftest
--selftest logs in, syncs once, and prints how many courses, assignments, materials and
grades it found. Get that working before wiring up the agent.
Register with Claude Code:
claude mcp add mcv -- uv --directory C:\Users\velys\Documents\mcp_mcv run mcv-mcp
Then ask: "What's due this week?", "Anything new on MCV since yesterday?", "Download the latest slides for Comp Prog."
Tools
| Tool | What it answers |
|---|---|
whats_new |
New/changed assignments, materials and grades since a point in time |
list_upcoming_deadlines |
What's due in the next N days, soonest first |
list_courses |
Enrolled courses, optionally one semester |
list_assignments |
Assignments with deadline + description |
get_assignment |
Full detail for one assignment |
list_materials |
Files posted in a course (folder, name, URL) |
download_material |
Save a file locally; extracts PDF text so it can be read |
list_grades |
Released scores |
sync_now |
Force a refresh (full=true for all semesters) |
auth_status |
Diagnostics: login state, last sync, row counts |
Configuration
| Variable | Default | Meaning |
|---|---|---|
MCV_USERNAME / MCV_PASSWORD |
— | Your MCV credentials |
MCV_LOGIN_METHOD |
itchula |
itchula for the IT Chula SSO button, mcv for the plain form |
MCV_POLL_MINUTES |
30 |
Poller interval (floor 10) |
MCV_POLL_ENABLED |
1 |
Set 0 to disable background polling |
MCV_STATE_DIR |
%LOCALAPPDATA%\mcv_mcp |
Where the DB, cookies and downloads live |
State is kept outside the repo. Cookies are persisted, so restarts don't re-login.
Run the poller separately (e.g. under Task Scheduler) with:
uv run python -m mcv_mcp.poller
Layout
src/mcv_mcp/
config.py settings, URLs, state paths
client.py Chula SSO login, cookie persistence, throttled AJAX transport
parsers.py HTML -> dicts (pure functions, fixture-tested)
store.py SQLite snapshot + event log
sync.py orchestration + background poller
server.py MCP tool definitions
MCV's markup is the fragile part, and it's confined to parsers.py. If something stops
returning rows, capture a fresh fixture (see docs/ENDPOINTS.md) and fix the selector there.
Status
Working against a real account: login, all four parsers, downloads and change detection are verified end to end. A representative sync pulled 9 courses, 20 assignments (every one with a parsed deadline), 112 materials and 69 graded items; the following sync reported zero changes, so the diff log stays quiet until something actually moves.
Known limits:
- Grade release is detected by polling, not pushed. A score appearing shows up as a
grade_changedevent within one poll interval (default 30 min). - Only the current semester is synced unless you pass
full=truetosync_now. - MCV's markup is the fragile part. If a parser starts returning zero rows, capture a fixture and fix the selector — see docs/ENDPOINTS.md.
Development
uv run pytest
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。