Coursera

The most advanced MCP server for Coursera.
_{279 tools · 5.6 MB binary · async Rust · zero dependencies · rich Obsidian export}

 ██████╗ ██████╗ ██╗   ██╗██████╗ ███████╗███████╗██████╗  █████╗       ███╗   ███╗ ██████╗██████╗
██╔════╝██╔═══██╗██║   ██║██╔══██╗██╔════╝██╔════╝██╔══██╗██╔══██╗      ████╗ ████║██╔════╝██╔══██╗
██║     ██║   ██║██║   ██║██████╔╝███████╗█████╗  ██████╔╝███████║█████╗██╔████╔██║██║     ██████╔╝
██║     ██║   ██║██║   ██║██╔══██╗╚════██║██╔══╝  ██╔══██╗██╔══██║╚════╝██║╚██╔╝██║██║     ██╔═══╝
╚██████╗╚██████╔╝╚██████╔╝██║  ██║███████║███████╗██║  ██║██║  ██║      ██║ ╚═╝ ██║╚██████╗██║
 ╚═════╝ ╚═════╝  ╚═════╝ ╚═╝  ╚═╝╚══════╝╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝      ╚═╝     ╚═╝ ╚═════╝╚═╝

Give any AI agent — Claude, GPT, or your own — full programmatic access to Coursera. Enroll in courses, speedrun lectures, solve quizzes, submit assignments, earn certificates, and push them to LinkedIn. All from a single MCP server.

Highlights

Python → Rust

Prerequisites

1. Install Cookie-Editor

All authentication uses browser session cookies. Install Cookie-Editor to export them:

Browser	Install
	Chrome Web Store
	Firefox Add-ons
	Edge Add-ons

2. Export Coursera Cookies

1. Log into coursera.org
2. Click Cookie-Editor icon → Export → copy CAUTH and CSRF3-Token values
3. Paste into .env (see Setup)

3. Export LinkedIn Cookies (optional)

1. Log into linkedin.com
2. Click Cookie-Editor icon → Export → copy li_at and JSESSIONID values
3. Paste into .env

Setup

git clone https://github.com/VoidChecksum/coursera-mcp.git
cd coursera-mcp

# Build the binary
cargo build --release

# Configure secrets
cp .env.example .env
chmod 600 .env

.env format

COURSERA_CAUTH=<paste CAUTH cookie value>
COURSERA_CSRF3=<paste CSRF3-Token cookie value>

LINKEDIN_LI_AT=<paste li_at cookie value>
LINKEDIN_JSESSIONID=<paste JSESSIONID value, without quotes>

# Optional: AI quiz solving via Gemini
GEMINI_API_KEY=<your Gemini API key>

Tuning parallelism

COURSERA_MCP_PARALLEL_COURSES=3   # concurrent course speedruns (default 3)
COURSERA_MCP_MAX_CONCURRENT=6     # concurrent Coursera API requests (default 6)
COURSERA_MCP_MAX_WORKERS=6        # concurrent lecture completions (default 6)
COURSERA_MCP_MAX_RETRIES=4        # retry count for transient failures (default 4)
COURSERA_MCP_BACKOFF_BASE=0.5     # backoff base seconds (default 0.5)

[!NOTE] When Coursera or LinkedIn logs you out, re-export cookies with Cookie-Editor and update .env.

Connect to Your AI Client

{
  "mcpServers": {
    "coursera": {
      "command": "/path/to/coursera-mcp/run.sh"
    }
  }
}

claude mcp add coursera /path/to/coursera-mcp/run.sh

Tools (279)

Course Discovery & Enrollment

Speedrunning

Progress, Grades & Certificates

Learner Analytics

Skill & Career Intelligence

Quiz & Assessment (GraphQL — no browser)

Peer Assignments (Native REST API — no browser)

Peer Review (Native REST API — no browser)

Discussion Forums (Native REST API — no browser)

Programming Assignments (Native REST API)

Content Extraction

Course Structure & Navigation

Item Completion (Pure REST/GraphQL)

HTTP Utilities

LinkedIn — Coursera Certificate Sync

LinkedIn — Profile & Identity

LinkedIn — Search

LinkedIn — Messaging

LinkedIn — Posts & Feed

LinkedIn — Network & Invitations

LinkedIn — Companies & Jobs

LinkedIn — Notifications & Analytics

LinkedIn — Groups

LinkedIn — Profile Management

Study & Planning

Obsidian Integration

Reviews & Ratings

Wishlist & Saved Courses

Announcements & Communications

Certificates & Accomplishments

Account & Profile Management

Session Management

Course Categories & Discovery

Partners & Instructors

Activity & History

Bookmarks

Labs & Workspaces

Degree Programs & Professional Certificates

Skills & Achievements

Batch Operations

Honor Code & Compliance

Financial Aid

Social & Community

Reverse Engineering

Several tools were built by intercepting live browser traffic and reverse-engineering Coursera's internal APIs:

Freemium Bypass — Coursera sets enrolledViaFreemiumAt on free enrollments, which causes all opencourse.v1 video-event and supplement-completion endpoints to return 403. The bypass deletes the enrollment via DELETE /api/openCourseMemberships.v1/{uid}~{courseId} (with an empty {} body — Akka backend requires it) and re-creates it via POST without the freemium flag.

GraphQL Gateway — 48+ operations captured from graphql-gateway, including GetLearnerStats, WeeklyStreaks_QueryWeeklyStreaks, GetLearnerWeeklyGoals, GetCourseBranchModuleProgressByCourse, GetLearnerCourseTimeline, GetMyCertificateByProduct, CourseNavigation, GetAiGradingEligibilityByCourse, CareerRolesQuery, and GetProductOwnership.

REST APIs — 61+ endpoints mapped from browser network traffic, including onDemandCourseViewGrades.v1, onDemandReferences.v1, onDemandLectureAssets.v1, onDemandInstructorNotes.v1, userNotes.v1, onDemandCourseForums.v1, onDemandLearningObjectives.v2, and subscriptions.v1.

Quiz Solver — Elimination-learning algorithm reimplemented natively in Rust (replacing the Python skipera library): submits answers, gets per-question feedback marking options correct/incorrect, persists learned answers to disk, and retries with improved answers until passing (≥70%). AI-powered first attempt via Gemini LLM, with systematic elimination as fallback. Handles START_NEW_ATTEMPT, RESUME_DRAFT, rate limits, option ID reshuffling, and all question types (multiple choice, checkbox, text reflect, rich text, numeric).

Architecture

coursera-mcp/
├── Cargo.toml           Rust project manifest
├── src/
│   ├── main.rs          Entry point — tokio runtime + MCP stdio transport
│   ├── config.rs        Environment-based configuration
│   ├── client.rs        Async HTTP client (retry, backoff, semaphore, caching)
│   ├── server.rs        ServerHandler — 279-tool dispatch + tool descriptors
│   └── tools/
│       ├── courses.rs       Course CRUD, search, specializations (9 tools)
│       ├── progress.rs      Grades, progress, stats, XP (21 tools)
│       ├── enrollment.rs    Enrollment, lecture completion, items (11 tools)
│       ├── quiz.rs          Full quiz solver engine (10 tools, 3500 lines)
│       ├── speedrun.rs      Speedrun, bulk solve (4 tools)
│       ├── content.rs       Transcripts, supplements, notes (14 tools)
│       ├── forum.rs         Forum threads, replies, upvotes (8 tools)
│       ├── assignments.rs   Peer, programming, written (22 tools)
│       ├── linkedin.rs      LinkedIn Voyager API (53 tools)
│       ├── obsidian.rs      Obsidian vault export (5 tools)
│       ├── dashboard.rs     Dashboard, deadlines, notifications (21 tools)
│       ├── ai_quiz.rs       Gemini LLM quiz solver (1 tool + helpers)
│       ├── automation.rs    Pure REST/GraphQL item completion (1 tool)
│       └── extra.rs         Reviews, wishlist, accounts, social, etc. (53 tools)
├── run.sh               Launcher (just execs the binary)
├── .env                 Your cookies (gitignored)
└── .env.example         Template

┌──────────────────────────────────────────────────────────────────────┐
│              MCP Clients (Claude / GPT / Custom)                     │
├──────────────────────────────────────────────────────────────────────┤
│            coursera-mcp  ·  rmcp stdio  ·  tokio async               │
├─────────┬─────────┬──────────┬──────────┬──────────┬────────────────┤
│  REST   │ GraphQL │  HTTP    │ LinkedIn │ Gemini   │ Obsidian       │
│  API    │ Gateway │  Client  │  Voyager │  LLM     │  API           │
│  (279   │  (50+   │ (reqwest │          │  API     │                │
│  endpts)│  ops)   │ +rustls) │          │          │  REST :27123   │
└─────────┴─────────┴──────────┴──────────┴──────────┴────────────────┘
       Single 5.6 MB binary — zero runtime dependencies

Security

[!CAUTION] Never commit your .env file. If you accidentally expose cookies, log out of Coursera/LinkedIn immediately to invalidate them.

License

MIT

_{Built for agents that learn faster than humans.}