penn-course-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@penn-course-mcp search for CIS-1200"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
penn-course-mcp
An MCP server for University of Pennsylvania course planning. It is built on the public Penn Courses API (the same backend behind Penn Course Plan and Penn Course Review), so the catalog, ratings, and schedule data are live.
With it, an MCP client (Claude Desktop, Claude Code, or any other) can search the course catalog, inspect sections and meeting times, read ratings and reviews, check a set of sections for time conflicts, assemble a weekly schedule, and save that schedule back to a Penn Course Plan account.
The server works with no configuration against the public API. Setting a session cookie additionally unlocks detailed Penn Course Review breakdowns and the schedule-write tools. It runs over stdio for local clients, or streamable HTTP when hosted.
Tools
Tool | What it does |
| Active term code (e.g. |
| Search the catalog by code or keyword; returns summaries and ratings |
| Full course info: description, prereqs, attributes, sections, meeting times |
| Sections with formatted meeting times and instructors; filter by status or activity |
| Public aggregate ratings (quality, difficulty, workload) |
| Detailed PCR reviews when authenticated; falls back to aggregates otherwise |
| Courses carrying a given attribute or Gen-Ed code |
| Courses fulfilling a requirement (by code or name) |
| Detect time conflicts among a set of section ids |
| Weekly grid, total credits, conflicts, and missing-companion warnings |
| Side-by-side ratings, difficulty, workload, and prereqs |
| Search and rank by quality, difficulty, or workload |
| List the saved schedules in your Penn Course Plan account (auth) |
| Write a schedule to your Penn Course Plan account so it shows on the site (auth) |
| Delete a saved Penn Course Plan schedule by id (auth) |
Course codes look like CIS-1200; section ids look like CIS-1200-001.
Related MCP server: Brown Courses MCP Server
Installation
uv pip install -e ".[dev]" # or: pip install -e ".[dev]"Running
# stdio (default), for Claude Desktop / Claude Code
penn-course-mcp
# streamable HTTP at http://127.0.0.1:8000/mcp
penn-course-mcp --transport http --port 8000
# development server / MCP Inspector
fastmcp dev src/penn_course_mcp/server.pyClient configuration
Add the server to your MCP client config (for Claude Desktop, this is
claude_desktop_config.json):
{
"mcpServers": {
"penn-course": {
"command": "uvx",
"args": ["penn-course-mcp"],
"env": {
"PENN_COURSES_SESSION_COOKIE": ""
}
}
}
}Use
"command": "penn-course-mcp"instead ofuvxif you installed the package into a virtualenv that is on yourPATH.
Configuration
All environment variables are optional (see .env.example):
Variable | Default | Purpose |
|
| API base URL |
|
| Default semester ( |
|
| Cache TTL (seconds) for catalog and detail data |
|
| HTTP timeout (seconds) |
|
| Sent with every request |
| (unset) | Enables detailed reviews and schedule writes (see below) |
|
|
|
|
| HTTP host |
|
| HTTP port |
Detailed reviews and schedule writes
The public API exposes aggregate ratings (course and instructor quality, difficulty, and work required) without authentication. The detailed per-instructor Penn Course Review breakdowns, and the schedule-write tools, require a logged-in Penn session. To enable them, log in to penncoursereview.com, copy your session cookie, and set:
export PENN_COURSES_SESSION_COOKIE="sessionid=...; csrftoken=..."get_course_reviews falls back to the public aggregate ratings when no cookie is set, so it
never errors. The schedule-write tools (list_schedules, save_schedule, delete_schedule)
read and modify a real Penn Course Plan account, so they always need the cookie. Writes also
need the csrftoken value, since Django enforces CSRF on unsafe methods. save_schedule
creates a new schedule by default; pass an existing schedule_id (from list_schedules) to
overwrite one instead.
The read-only catalog and rating tools never modify your account. Only the three schedule tools do.
Claude skill
This repository ships an optional Agent Skill
at .claude/skills/penn-course-planning. It
gives Claude guidance on using the tools well, including normalizing course codes (CIS 120
to CIS-1200), reading the rating scales, and choosing the right tool for schedule and
requirement questions. It activates automatically when you use Claude Code in this repository.
To make it available everywhere, copy it into your user skills folder:
cp -R .claude/skills/penn-course-planning ~/.claude/skills/Development
uv pip install -e ".[dev]"
pytest # pure planning logic, mocked client, and tool tests
flake8 # style, import order (isort), and quote checks
black . # formatterThe planning logic in src/penn_course_mcp/planning.py is pure and network-free, so the
schedule-conflict detection, time formatting, and comparison are covered by unit tests
without hitting the API.
Notes
Meeting times use Penn's
HH.MMencoding (10.15is 10:15,15.3is 15:30). The server renders them as readableHH:MMranges.Day codes are
M,T,W,R,F, whereRis Thursday.Back-to-back meetings, where one ends exactly as the next begins, are not counted as conflicts.
This is an unofficial tool that consumes the public Penn Courses API. It caches responses and caps concurrency; please use it respectfully.
License
MIT
Available Tools
15 toolsbuild_scheduleA
Assemble a weekly schedule from section ids: total credits, a day-by-day grid, time conflicts, and warnings for missing required companion sections.
Inputs: sections (required), semester. Returns: object
check_schedule_conflictsA
Check a set of section ids (e.g. ['CIS-1200-001','MATH-1400-002']) for time conflicts. Returns the clashing meetings in human-readable form.
Inputs: sections (required), semester. Returns: object
compare_coursesB
Compare courses side by side on ratings, difficulty, workload, credits, and prerequisites. Accepts a list of course codes (e.g. ['CIS-1200','CIS-1600']).
Inputs: course_codes (required), semester. Returns: object
delete_scheduleA
Delete a saved Penn Course Plan schedule by its id (requires auth).
Use list_schedules to find the id first. This permanently removes the
schedule from your Penn Course Plan account.
Inputs: schedule_id (required). Returns: object
find_courses_by_attributeA
Find courses carrying a given attribute (e.g. 'EUMS', 'MFR', a Gen-Ed code).
match='code' requires an exact attribute-code match; match='description'
does a substring search across codes and descriptions. Supply prefilter_query
to narrow the candidate set first (faster, smaller payload).
Inputs: attribute (required), limit, match, semester, +1 more. Returns: object
find_courses_by_requirementB
Find courses fulfilling a requirement (pre-NGSS requirement code or name).
Substring-matches against course attributes and pre-NGSS requirement codes/names.
Inputs: requirement (required), limit, semester. Returns: object
get_course_detailsC
Get full details for a course: description, prerequisites, attributes, crosslistings, aggregate ratings, and all sections with meeting times.
Inputs: course_code (required), semester. Returns: object
get_course_ratingsB
Return the public aggregate ratings for a course (always available).
Includes course_quality, instructor_quality, difficulty, and work_required.
Inputs: course_code (required), semester. Returns: object
get_course_reviewsA
Get detailed Penn Course Review breakdowns for a course.
When an authenticated session cookie is configured this returns the full per-instructor review data. Otherwise it gracefully falls back to the public aggregate ratings and reports an unauthenticated status (never errors).
Inputs: course_code (required). Returns: object
get_current_semesterA
Return the active semester and whether detailed reviews are enabled.
Useful as a first call to learn the current term code (e.g. '2026C') and the server's review-auth status.
Inputs: no named parameters. Returns: object
list_course_sectionsA
List a course's sections with formatted meeting times and instructors.
open_only keeps only sections whose status is open; activity filters by
type (e.g. 'LEC', 'REC', 'LAB').
Inputs: course_code (required), activity, semester, open_only. Returns: object
list_schedulesA
List the saved schedules in your Penn Course Plan account (requires auth).
Returns each schedule's id, name, semester, and the section ids it contains —
handy for finding a schedule_id to update or delete.
Inputs: no named parameters. Returns: object
recommend_coursesA
Search and rank courses by a quality metric.
sort_by is one of: 'course_quality', 'instructor_quality',
'recommendation_score' (higher is better), or 'difficulty' / 'work_required'
(lower is better, for finding lighter courses).
Inputs: query (required), limit, sort_by, semester. Returns: object
save_scheduleA
Save a schedule to your Penn Course Plan account so it appears on the website.
Creates a new schedule named name containing the given sections (full
section ids like 'CIS-1200-001'). Pass an existing schedule_id to overwrite
that schedule instead of creating a new one. Writes to your live Penn account
and requires PENN_COURSES_SESSION_COOKIE (sessionid + csrftoken).
Inputs: name (required), sections (required), semester, schedule_id. Returns: object
search_coursesA
Search the course catalog by code or keyword (e.g. 'CIS-1200' or 'ethics').
Returns lightweight course summaries including aggregate ratings. semester
defaults to the current term.
Inputs: query (required), limit, semester. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 15 tools
Each tool has a distinct purpose: searching, retrieving details, listing sections, getting ratings/reviews, finding by attribute/requirement, checking conflicts, building schedules, comparing, recommending, and managing saved schedules. No overlapping functionality.
All tool names follow a consistent verb_noun pattern using snake_case (e.g., search_courses, get_course_details, build_schedule). The verbs vary (search, get, list, find, check, build, compare, recommend, save, delete) but are descriptive and predictable.
15 tools cover the full lifecycle of course exploration and schedule planning without being excessive. Each tool serves a clear purpose in the domain, and the count is appropriate for a comprehensive course management MCP.
The tool surface covers all major operations: course search, details, ratings, reviews, attribute/requirement filtering, schedule building, conflict checking, comparison, recommendations, and persistent schedule management. No obvious gaps for the stated purpose.
Maintenance
Related MCP Connectors
Unofficial Wentworth (WIT) course catalog and student schedule. Needs a WIT sign-in.
Manage your Canvas coursework with quick access to courses, assignments, and grades. Track upcomin…
Unofficial NTNU course data: search, timetables, grades, course info, and exam logistics.
Study planner for high school students. Manage tasks, deadlines, and schedules.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables searching and retrieving course details, schedules, prerequisites, and instructor information from the CMU course catalog via the ScottyLabs API.92MIT
- FlicenseAqualityDmaintenanceEnables searching Brown University courses, getting detailed course info, and checking schedule conflicts via natural language, using the university's public course catalog API.5-
- FlicenseAqualityDmaintenanceLocal-first Rutgers course planning assistant that answers 'what should I take and when?' using a CP-SAT solver, real transcript data, and live SOC data, connected to Claude via MCP.13-
- AlicenseBqualityBmaintenanceEnables users to query official undergraduate and graduate course schedules, indexes, prerequisites, and exam timetables, including tools for comparing indexes and finding matching schedules. It is read-only, local, and supports filtering by days, times, lesson types, teaching weeks, and schedule clashes.71,874 PyPI1MIT