Skip to main content
Glama
deenesjakoruzh
unielevotor

nwafu-mcp

by unielevotor
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

nwafu-mcp · Northwest A&F University Campus Information MCP Toolkit

Provides Agents deployed on any agent platform with a set of tools compliant with the MCP protocol, covering three core capabilities:

  1. Campus channel hot summary: Fetches recent hot posts from the official QQ channel of Northwest A&F University (pd.qq.com/g/inwafu1934), ranks them by interaction count, and automatically categorizes them into "Activities / Competitions / Notices / Recommendations / Tips / Help" and other categories; every important item is accompanied by the source post title and link, ensuring traceability.

  2. Official website quick query: Quickly looks up recent activities, competitions, notices, and recruitment information through the full-text indexes of the official website (www.nwafu.edu.cn) and the news network; keywords such as "College of Plant Protection" or "Academic Affairs Office" can be passed in to narrow the search scope.

  3. Cross-site custom search: Automatically splits the user's question into search terms, searches both the official website full-text index and recent campus channel posts, then merges and outputs the results; all results are formatted as readable Markdown with links to the original sources.

Features

  • Standard MCP implementation: Based on the official mcp Python SDK (FastMCP) with stdio transport, ready to connect directly to Claude Desktop, Cursor, Windsurf, and any MCP-compatible agent platform.

  • Traceable information: Every conclusion item carries the source title and original link; important information is given a dedicated "Key Reminder / Timeliness Reminder" section.

  • Smart categorization: The rule engine automatically classifies channel posts into "Recommendations / Tips / Activities / Competitions / Notices / Help" and other categories, sorted within each group by popularity (likes + weighted comments).

  • Precise keyword search: Official website queries support any keyword to narrow the scope (e.g., college name, department name, or item name).

  • Robust fault tolerance: Request timeouts, retries, randomized rate-limit delays, and Cookie expiry warnings are built in; a failure of one tool does not affect other tools.

  • Privacy compliance: Only publicly visible content is fetched, without bypassing login/permission checks; Cookies are injected via environment variables and never written into the code repository.

Related MCP server: who-will-notify-mcp

Tool List

Tool

Purpose

Key Parameters

campus_channel_summary

Smart summary of recent hot posts on the campus QQ channel (hot ranking + categorization + key information)

window_hours time window, max_posts fetch volume, top_n number of hot-ranking entries, include_comments whether to fetch hot comments

official_site_recent

Quick query and summary of recent notices/activities/competitions/recruitment on the official website

category category, keyword to narrow the scope (e.g., college name), days time range, max_results

official_site_search

Full-text search of the official website (custom keywords)

query core term, keyword additional qualifier, days, max_results

campus_question_search

Cross-site custom question search over "official website + campus channel"

question original question text, keywords optional explicit search terms, days, include_channel

All tools output Markdown text with the structure: data source, fetch time, query conditions, result list (title / date / source / summary / link), and key reminders.

How It Works

Campus QQ Channel

The Tencent Channel web client is purely front-end rendered; its public data comes from a protobuf-over-JSON gateway:

POST https://pd.qq.com/qunng/guild/gotrpc/noauth/trpc.qchannel.commreader.ComReader/<Method>
x-oidb: {"uint32_service_type": 11 时间线 | 5 评论}

Public content is anonymously readable, but the gateway requires a browser-level session Cookie (p_uin + uuid + EO-Bot-Js-Token). This tool:

  • By default, pulls the latest N posts (by publish time) from the "Post Square" sub-channel (670126629);

  • Estimates popularity with 点赞数 + 2 × 评论数 and picks the hottest posts within the window;

  • Automatically categorizes posts with a rule engine and separately highlights important information (notices/announcements/registration/deadlines/exams, etc.);

  • Optionally fetches hot comments for hot-ranking posts (GetFeedComments).

Note: The "hot sort" endpoint of GetGuildFeeds returned empty data in real-world testing, so the hot ranking uses the "recent posts × interaction count" approach, which better matches the meaning of "recently hot".

University Official Website

The university's main site and news network use 通元 CMS full-text search:

GET https://www.nwsuaf.edu.cn/cms/web/search/index.jsp
    ?query=<关键词>&siteID=<站点ID>&searchScope=0&channelID=&matchType=0
    &sortField=publishDate&order=1&date=<3|6|12>&page=<页码>

  • siteID=32e6d9be... is the main-site index (covering all colleges/departments of the university);

  • Search terms are generated by category (notices/announcements, activities/lectures/forums, competitions/contests, recruitment, etc.), merged with user keywords, and fetched page by page, parsing title, date, source, summary, and original link;

  • The default time range is the recent 90 days (mapped to the index's 3/6/12-month filters).

Quick Start (Local Run)

cd nwafu-mcp
uv sync --extra dev

1. Configure the Channel Cookie (Optional, Only Needed for Channel Tools)

uv run nwafu-export-cookies --out cookies.json

The script opens the channel page in your local Edge/Chrome to establish a session and exports the Cookie. Then write the cookie_header from cookies.json into an environment variable:

$env:PDQQ_COOKIES = "p_uin=xxx; uuid=xxx; EO-Bot-Js-Token=xxx"
# 或
$env:NWAFU_COOKIE_FILE = "F:\path\to\cookies.json"

Cookies expire (the anti-crawler token is bound to the browser instance); for cloud deployment, refresh them regularly.

2. Test the Tool Functions Directly (Without an MCP Client)

uv run python -c "
import os
os.environ['NWAFU_COOKIE_FILE'] = 'cookies.json'
from nwafu_mcp.server import official_site_recent
print(official_site_recent(category='通知', keyword='植保学院'))
"

3. Run as an MCP Server

uv run nwafu-mcp

Connecting to MCP Clients

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "nwafu-campus": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/unielevotor/nwafu-AgentPlatformMCP",
        "nwafu-mcp"
      ],
      "env": {
        "PDQQ_COOKIES": "p_uin=xxx; uuid=xxx; EO-Bot-Js-Token=xxx"
      }
    }
  }
}

Cursor / Windsurf / General Platforms (JSON Form)

{
  "mcpServers": {
    "nwafu-campus": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/unielevotor/nwafu-AgentPlatformMCP", "nwafu-mcp"],
      "env": {
        "PDQQ_COOKIES": "p_uin=xxx; uuid=xxx; EO-Bot-Js-Token=xxx",
        "PDQQ_GUILD_ID": "inwafu1934",
        "PDQQ_CHANNEL_ID": "670126629"
      }
    }
  }
}

It can also be installed with pip install git+https://github.com/unielevotor/nwafu-AgentPlatformMCP and then started directly with the nwafu-mcp command.

GitHub Deployment Steps

  1. Push to the GitHub repository (unielevotor/nwafu-AgentPlatformMCP):

git remote add origin https://github.com/unielevotor/nwafu-AgentPlatformMCP.git
git push -u origin main

  1. Make sure the repository is Public (for a private repository, the target agent platform must be able to access GitHub credentials).

  2. In the agent platform's MCP configuration, fill in the uvx --from git+... startup command above and inject environment variables such as PDQQ_COOKIES.

Environment Variables

Variable

Required

Description

PDQQ_COOKIES

Required for channel tools

QQ channel browser session Cookie

NWAFU_COOKIE_FILE

Required for channel tools

Points to the JSON exported by nwafu-export-cookies

PDQQ_GUILD_ID

No

Channel identifier, default inwafu1934

PDQQ_CHANNEL_ID

No

Sub-channel ID, default 670126629 (Post Square)

PDQQ_MIN_DELAY / PDQQ_MAX_DELAY

No

Channel request interval (seconds), default 0.3–0.8

NWAFU_TIMEOUT

No

Per-request timeout (seconds), default 30

See .env.example.

Sample Output (Excerpt)

# 🎓 西农校园频道 · 近期热门总结

> 数据来源：西北农林科技大学官方 QQ 频道 ｜ 时间窗口：近 72 小时 ｜ 帖子数：48

## 🔥 热度榜
1. **中元节鬼是真的多啊…**（👍7 · 💬1 · 08-27）
   来源：[中元节鬼是真的多啊](https://pd.qq.com/g/inwafu1934/post/...)

## 🏆 竞赛（2）
- **关于举办2026年创新创业大赛的通知**（👍5 · 💬2 · 08-26）
  来源：[…](https://pd.qq.com/g/inwafu1934/post/...)

## ⚠️ 重点信息（建议优先查看）
- [关于选课时间安排的通知](https://pd.qq.com/g/inwafu1934/post/...)
# 📢 西北农林科技大学官网 · 近期信息查询

> 查询条件：分类=通知 ｜ 关键词=植保学院 ｜ 时间范围：近 90 天 ｜ 结果数：12

1. **关于举办2026年植保论坛系列学术报告会（十七）的通知**（2026-08-26 · 植物保护学院）
   [查看原文](https://ppc.nwafu.edu.cn/xzbg/...)

Compliance and Usage Recommendations

  • Only fetch publicly visible content; never bypass login, paywall, or permission checks.

  • Random UA, random delays, and failure retries are built in. Do not reduce the delays for high-frequency scraping when deploying, and comply with the target platform's terms of service and applicable regulations such as the Personal Information Protection Law.

  • For important information (registration deadlines, exam arrangements, etc.), always refer to the official source; tool output is for quick reference only.

  • Channel Cookies contain session identifiers; never commit them to a public repository — always inject them via environment variables.

FAQ

Symptom

Resolution

Channel tool reports "Cookie not configured"

Run nwafu-export-cookies --out cookies.json locally, then configure the environment variables

Channel tool reports retcode=150/4000

Cookie expired or anti-crawler token missing; re-export them

Official website query returns empty results

Widen days, remove the keyword, or switch category, then retry

No Edge/Chrome in the cloud

Install playwright and run python -m playwright install chromium, or periodically export the Cookie locally and inject it

Directory Structure

src/nwafu_mcp/
  server.py          MCP 服务器与四个工具定义
  qq_channel.py      QQ 频道数据层（时间线/热评）
  official_site.py   官网全文检索数据层
  classify.py        帖子分类与热度评分
  format.py          Markdown 报告排版
  config.py          环境变量与默认配置
  export_cookies.py  本地导出频道 Cookie 的 CLI
tests/               单元测试
scripts/mcp_smoke.py 端到端 MCP 冒烟测试（连接 stdio server 并调用全部工具）

Available Tools

4 tools
campus_channel_summaryA

抓取西北农林科技大学官方 QQ 频道（https://pd.qq.com/g/inwafu1934）近期帖子，按互动量（点赞+评论）排名出热度榜，并自动归类为活动/竞赛/通知/推荐/贴士/求助/其他，输出 Markdown 总结；每条重要信息都附来源帖子标题与链接。

ParametersJSON Schema
NameRequiredDescriptionDefault
top_nNo
max_postsNo
window_hoursNo
comment_top_nNo
include_commentsNo

Output Schema

ParametersJSON Schema
NameRequiredDescription
resultYes

TDQS

A3.7/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description carries the full burden. It transparently discloses the source URL, ranking formula, automatic categorization, Markdown output, and source attribution. It does not mention failure modes or rate limits, but the core behavior is clearly described.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single dense sentence that front-loads the fetch action, source, ranking, categorization, and output format. Every clause is relevant, though a more structured breakdown of parameters would improve readability.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The output schema covers return structure, and the description explains source, ranking, categories, and citations. However, with five parameters and no explicit guidance on when to invoke this tool versus siblings, the description is not fully complete.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, and the description does not explain any of the five parameters. Terms like 'top_n' and 'window_hours' are only indirectly related to '热度榜' and '近期帖子', while max_posts, comment_top_n, and include_comments receive no clarification.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific action: fetch recent posts from a named QQ channel, rank them by likes+comments, categorize them into seven types, and output a Markdown summary with source links. It clearly identifies the resource and unique behavior, distinguishing it from a plain recent-posts or search tool.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description implies this tool is for obtaining a ranked, categorized summary of QQ-channel posts rather than raw retrieval or search. It does not explicitly state when to use this tool versus the sibling tools or describe exclusions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

official_site_recentA

查询西北农林科技大学官网（https://www.nwafu.edu.cn/）最近的活动、竞赛、通知、招聘等官方信息并输出 Markdown 总结。支持传入关键词（如学院名）缩小检索范围，例如 keyword='植保学院' 只返回植保学院相关通知。

ParametersJSON Schema
NameRequiredDescriptionDefault
daysNo
keywordNo
categoryNo全部
max_resultsNo

Output Schema

ParametersJSON Schema
NameRequiredDescription
resultYes

TDQS

A3.5/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the behavioral burden; it discloses that this is a read-only query of the official site and that output is a Markdown summary, plus that keyword narrows the results. It does not mention rate limits, freshness boundaries, or failure modes, but for a simple read tool the key behavioral traits are disclosed.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two sentences with no filler: the first establishes tool, source URL, content scope, and output format; the second gives the key parameter example. The most load-bearing information is front-loaded.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The tool is simple and the output schema is present, which covers return structure. However, the description does not explain three of the four parameters, nor does it position the tool against official_site_search, so an agent still has meaningful uncertainty about how to set days/category/max_results and when to choose this over its siblings.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, so the description must compensate, but only keyword is given real meaning (with the '植保学院' example). days, category, and max_results are left to inference from their names and defaults, which is insufficient with no schema descriptions.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific verb–resource pair: query the NWAFU official website for recent activities, competitions, notices, and recruitment information, then output a Markdown summary. It does not explicitly distinguish itself from official_site_search or the other siblings, so it loses a point for lacking sibling differentiation.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The intended use—retrieving recent official-site news—is implied and there is a concrete keyword example, but no alternative tools are named and no when-not-to-use conditions are given. Sibling tools like official_site_search could overlap, yet the description does not route the agent between them.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 4 tool updatesv0.1.0
    • First observedcampus_channel_summary
    • First observedcampus_question_search
    • First observedofficial_site_recent
    • First observedofficial_site_search
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A3.7/5.0

Scored across 4 tools

Disambiguation4/5

Each tool targets a fairly distinct retrieval need: QQ channel hot summaries, official-site recent info, official-site full-text search, and cross-source question search. The only mild overlap is between official_site_recent and official_site_search, but their descriptions make the difference clear enough.

Naming Consistency4/5

Tool names are consistently lowercase snake_case and follow recognizable source prefixes like campus_* and official_site_*. However, suffixes mix result-oriented terms (summary, recent) with action terms (search), so they do not follow a strict verb_noun pattern throughout.

Tool Count5/5

Four tools is well-scoped for a campus information retrieval server. Each tool covers a distinct retrieval workflow without redundancy, and the set is neither too thin nor overloaded for its stated purpose.

Completeness4/5

The set covers recent official announcements, full-text official-site search, QQ channel hot content, and a combined question-based search across both sources. A dedicated channel-only keyword search is missing, but campus_question_search partially fills that gap, so the surface is reasonably complete.

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • cc98-mcp

    EviterLesRoses2
    A
    license
    A
    quality
    D
    maintenance
    cc98-mcp enables AI assistants to search, read, and aggregate posts from the Zhejiang University campus forum CC98, using official read-only API tools.
    10
    3
    MIT

  • zju-mcp

    EviterLesRoses2
    F
    license
    A
    quality
    C
    maintenance
    Enables 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.
    7
    3
    -

  • google-surf-mcp

    tamir-ariunsukh
    A
    license
    A
    quality
    B
    maintenance
    Provides Google search, URL extraction, and academic paper inline extraction without API keys, enabling search and content retrieval from a single MCP server.
    5
    1,307 npm
    MIT