chaoxing-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., "@chaoxing-mcp List my courses and check for new assignments"
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.
cxmcp — 学习通 MCP Server(教师版 / 学生版)+ 配套 Skills
参考 Cr-GH0/chaoxing-agent 与 Aemiser/chaoxing_mcp 的逆向成果自研的 超星学习通 MCP server。纯 HTTP 协议模拟,无浏览器、无 WebDriver、无油猴依赖。
⚠️ 使用声明
本项目为学习通平台第三方协议实现,平台改版可能导致功能失效;
视频/测验自动化能力默认关闭,开启后可能违反平台条款与所在院校学术规范, 存在成绩无效、账号处罚风险,后果自负;
密码仅在登录调用内存中使用;Cookie 以 0600 权限存本地 JSON。
Related MCP server: zju-mcp
组成
目录 | 说明 |
| MCP server 包,双入口 |
| WorkBuddy/CodeBuddy 教师版 Skill 插件 |
| WorkBuddy/CodeBuddy 学生版 Skill 插件 |
工具规模:共用 5 + 教师 12 + 学生 11;学生版加 --enable-automation 追加 2 个自动化工具。
安装
cd chaoxing-mcp
pip install -e . # 或 uv pip install -e .
cx-mcp-student --help # 验证入口接入宿主(stdio,即插即用)
{
"mcpServers": {
"chaoxing-teacher": {
"command": "cx-mcp-teacher",
"env": { "CX_COOKIE_FILE": "~/.chaoxing-mcp/teacher_cookies.json" }
},
"chaoxing-student": {
"command": "cx-mcp-student",
"env": { "CX_COOKIE_FILE": "~/.chaoxing-mcp/student_cookies.json" }
}
}
}远程模式:
cx-mcp-student --transport streamable-http --host 127.0.0.1 --port 8000(不要绑定公网)自动化(学生版):
cx-mcp-student --enable-automation或环境变量CX_ENABLE_AUTOMATION=1
登录方式
账号密码:
cx_login(登录页t=true时走 AES-CBC 加密提交/fanyalogin)。 平台要求短信/二次验证时返回login_verification_required,不绕过。Cookie 导入:浏览器登录学习通 → F12 → 复制 Cookie(
uid/_uid/fid/vc3等)→cx_import_cookies传入[{name, value, domain, path}]。
安全设计(借鉴 chaoxing-agent)
两阶段确认:批改/发通知/发起签到/上传/正式提交/自动化操作,首次调用返回
confirmation_required+ 一次性令牌(SHA-256 参数指纹绑定,5 分钟 TTL), 参数变化即失效,防重放、防 TOCTOU;风控即暂停:HTTP 202 / antispider 跳转 → 返回
risk_control_paused与指引, 不自动过验证码;回读验证:提交类操作以刷新列表核对状态为准,HTTP 200 不等于成功;
日志全部走 stderr,stdout 仅承载 MCP stdio 协议。
实机校准(重要)
本仓库协议实现基于参考项目的公开逆向知识 + mock 单测(23 个,全绿), 未在真实学习通账号上验证。上线前请先跑:
python smoke_test.py login --role student --username 手机号 --password 密码登录与课程列表属高确定性端点;签到/成绩/通知/上传/自动化端点随平台模板变动,
失败时会返回候选端点探测记录,按记录抓包更新 src/cxmcp/api/ 中对应 URL 即可。
开发
pip install -e ".[dev]"
pytest tests/ -q目录
src/cxmcp/
├── config.py # 环境变量/CLI 配置
├── crypto.py # 登录 AES-CBC(固定 TRANSFER_KEY)
├── session.py # 会话/Cookie 原子持久化/TLS 兼容/风控守卫
├── confirm.py # 两阶段确认令牌
├── api/ # courses / chapters / work / files / exams / signin / teacher / automation
├── tools/ # MCP 工具注册(common / teacher / student / automation)
├── server.py # build_server(role, enable_automation)
└── cli.py # stdio / streamable-http 双传输入口许可
MIT。与超星官方无关联。
Available Tools
16 toolscx_import_cookiesA
从浏览器导入学习通 Cookie(登录后 F12 复制,或用编辑器插件导出)。传入 [{name,value,domain?,path?}, ...]。适用于平台强制二次验证、密码登录被风控等场景。
Inputs: cookies (required). Returns: object
cx_list_coursesA
列出学习通课程(学生返回『我学的课』,教师返回任教课程),含 courseId/clazzId/cpi——后续所有课程相关操作都靠这三个 ID 定位。
Inputs: no named parameters. Returns: object
cx_loginA
登录学习通(超星账号)。密码仅在本次调用内存中使用,不会写入日志或返回值。若平台要求二次验证(短信/扫码),会返回 login_verification_required,此时请引导用户浏览器登录后改用 cx_import_cookies。
Inputs: password (required), username (required). Returns: object
cx_logoutA
退出学习通并删除本地 Cookie 文件。
Inputs: no named parameters. Returns: object
cx_s_download_attachmentC
下载作业/资料附件到本地(objectId 定位)。
Inputs: saveDir (required), objectId (required). Returns: object
cx_s_download_materialC
按 objectId 下载课程资料(getYunFiles 直链优先)。
Inputs: saveDir (required), objectId (required). Returns: object
cx_session_statusA
检查学习通登录态是否有效(回读个人空间首页判断)。
Inputs: no named parameters. Returns: object
cx_s_get_homework_detailA
读取作业详情:题目列表(题干/选项/题型)、附件列表、standardEnc。不会自动作答或提交。
Inputs: workUrl (required), courseRef (required). Returns: object
cx_s_get_study_recordsC
查看课程学习记录(菜单『学习记录/统计』页解析)。
Inputs: courseRef (required). Returns: object
cx_s_list_chaptersB
列出课程章节树(章节名/链接)。
Inputs: courseRef (required). Returns: object
cx_s_list_examsA
列出课程考试(标题/状态/入口链接)。只读取,不进入答题。
Inputs: courseRef (required). Returns: object
cx_s_list_homeworksB
列出课程作业(标题/状态/时间/详情链接)。
Inputs: courseRef (required). Returns: object
cx_s_list_learning_coursesA
列出『我学的课』(courseId/clazzId/cpi/教师名),后续操作用这些 ID 定位。
Inputs: no named parameters. Returns: object
cx_s_list_materialsA
列出课程『资料』目录的文件(名称/objectId)。
Inputs: courseRef (required). Returns: object
cx_s_sign_inA
学生签到:normal(一键)/location(位置)/code(签到码)/photo(拍照)。
Inputs: activityId (required), code, address, latitude, +3 more. Returns: object
cx_s_submit_homeworkA
作业作答:submit=false 为暂存(安全);submit=true 为正式提交(★高影响,首次调用返回 confirmToken,确认后携带 token + 同参数重调)。answers 格式: [{"stem":"题干关键词","answer":"A"}]。
Inputs: answers (required), workUrl (required), courseRef (required), submit, +1 more. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 16 tools
Most tools have clearly scoped purposes, but cx_s_list_learning_courses and cx_list_courses overlap heavily for student course listing, and cx_s_download_material vs cx_s_download_attachment both perform objectId-based downloads from different sources. Descriptions help clarify the intended resource, but an agent could still misselect between these pairs.
The set largely follows a verb_noun pattern with a consistent cx_s_ prefix for student operations and cx_ for auth/session operations. Minor inconsistencies exist: cx_session_status is noun-style, cx_s_list_learning_courses vs cx_list_courses use different names for similar listing actions, and material/materials singular/plural are mixed.
16 tools is at the upper edge but still reasonable for a platform server covering auth, courses, chapters, materials, homework, exams, sign-in, and downloads. There is some redundancy that could be trimmed, but no tool feels entirely out of place.
The server covers the main student workflows: login/session management, course listing, content/material access, homework reading and submission, sign-in, and exam listing. The most notable gap is that exams are read-only with no submission workflow, but the core learning-assistant surface is otherwise coherent.
Related MCP Connectors
MCP server for AI dialogue using various LLM models via AceDataCloud
Connect AI to your flomo notes. Search, create, edit notes and manage tags via MCP.
Connect any AI agent to 1,000+ apps and 27,000+ actions through one remote MCP server (OAuth).
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to interact with Yuque (语雀) platform through MCP protocol, supporting knowledge base management, document operations, search, and team collaboration features.11MIT
- FlicenseAqualityDmaintenanceEnables interaction with Zhejiang University's learning platform (学在浙大 / 智云课堂) via MCP tools, allowing natural language commands to check todos, view schedules, fetch lecture transcripts, and submit homework.79-
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to read and manipulate ZSXQ (知识星球) content via MCP, allowing listing groups, searching posts, downloading and summarizing content through natural language.9MIT
- AlicenseBqualityDmaintenanceEnables AI clients to operate Feishu/Lark through MCP protocol, including sending messages, managing chats, accessing contacts, and checking calendar.8MIT