osiris-mcp-server
Exposes Dutch university Osiris course catalogs as MCP tools, enabling search, retrieval, and exploration of course information across multiple institutions via the public Osiris Student API.
README
osiris-mcp-server
An MCP server exposing Dutch university Osiris course catalogs as tools. It
reads from the public, read-only Osiris Student API that universities publish
at https://<host>.osiris-student.nl/ (e.g. TU/e, TU Delft, Radboud,
Twente, Wageningen).
Quick start
Requires Python 3.11+ and uv.
uv sync
uv run osiris # run the MCP server over stdio
The server targets TU/e (tue) by default and accepts an institution code
per call, e.g. tue, tudelft, rug, ut:
search_courses("4DM00", institution="tudelft")
get_course("4DM00", institution="tue")
Usage with MCP clients
Add the server to your MCP client configuration.
{
"mcpServers": {
"osiris": {
"command": "uv",
"args": ["run", "--directory", "/path/to/osiris-mcp-server", "osiris"]
}
}
}
For clients that support environment variables (e.g. OSIRIS_LANGUAGE,
OSIRIS_TIMEOUT), pass them via the env key.
Tools
| Tool | Description |
|---|---|
list_faculties() |
Faculties/departments, code + name |
list_course_types() |
Course types (Bachelor College, Graduate School, ...) |
get_available_years() |
Academic years in the catalog |
get_categories() |
Course categories/levels |
get_blocks() |
Period/block tokens (e.g. 1, GS1, JAAR) |
get_timeslots() |
Timeslots (e.g. A, B1) |
get_units() |
Coordinating units (capaciteitsgroep/sectie) |
get_lecturers() |
Lecturer names |
get_languages() |
Instruction languages |
get_education_types() |
Education types (e.g. Challenge based) |
get_course_structure() |
Sections/fields a course contains |
list_courses(faculty, course_type, ...) |
{code, id, name, year} list for a faculty + type |
get_course(code, year) |
Course header (id, code, name, ECTS, course_type, owner, level, year, blocks) |
get_course_details(code) |
Full detail: content, exams, lecturers, materials, deep link, ... |
get_course_details_by_id(id) |
Full detail by internal course id |
search_courses(keyword, ...) |
Full-text search with filters |
search_minors(keyword, ...) |
Search minors (coherent packages, certificates) |
get_minor_details(code) |
Full minor detail: study programme, entrance requirements, ... |
get_minor_details_by_id(id) |
Full minor detail by internal minor id |
The list_courses and search_courses tools accept the same filters; every
one is optional except faculty for list_courses:
| Filter | Description |
|---|---|
faculty |
Faculty code (e.g. EE) or full name |
course_type |
Course-type code or full name (default BC) |
category |
Category/level (e.g. Introductory, Advanced) |
year |
Start year (2025) or range (2025-2026) |
period |
Block token (e.g. 1 for bachelor Q1, GS1 for graduate Q1) |
timeslot |
Timeslot code, prefix match (e.g. A matches A1, A2, ...) |
unit |
Coordinating unit (capaciteitsgroep/sectie) |
lecturer |
Lecturer name |
language |
Instruction language (name or 2-letter code) |
education_type |
Education type (e.g. Challenge based) |
available |
Only courses with places available (true) or not (false) |
Search results include a blocks list per course with the offering periods:
block token, dates, timeslots, location and availability — plus lecturers,
education_type, faculty and coordinating unit.
search_minors accepts the same year, faculty and education_type filters;
results include startmoments (academic year + block the minor can be started)
and study_programs (degree programmes the minor serves).
Configuration
| Variable | Default | Purpose |
|---|---|---|
OSIRIS_RELEASE_VERSION |
(unset) | Optional search-index header |
OSIRIS_MANIFEST |
(unset) | Optional search-index header |
OSIRIS_LANGUAGE |
EN |
Content language (NL or EN) |
OSIRIS_TIMEOUT |
30 |
HTTP timeout in seconds |
Most institutions serve a sensible default index without OSIRIS_RELEASE_VERSION
or OSIRIS_MANIFEST. Only set them to pin a specific Elasticsearch index
version (e.g. one extracted from a university's Osiris app bundle).
Development
uv run pytest # unit tests (offline)
uv run pytest -m e2e # live tests against the configured catalog (default TU/e)
uv run ruff check .
uv run ruff format --check .
uv run mypy src
License
The code is MIT licensed; see LICENSE.
The course data comes from the relevant university's Osiris API and is owned by that institution. This project is not affiliated with or endorsed by any university, and the MIT license does not cover the underlying course data.
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。