meeting-summary-mcp · Universal Meeting Summary MCP

A universal MCP Server: hand it any meeting (Feishu Minutes/Meeting Notes link, date range, or transcripts exported from Tencent Meeting, etc.), and it will automatically organize it into a standard meeting summary — Meeting Info / Meeting Summary / Meeting Content / To-Dos · Distributed by Department.

Use case: after a meeting → let AI read the meeting content → automatically produce a structured summary, and assign each to-do to the corresponding department ("Product Dept. / Operations Dept. / Brand Dept. …"), saving the manual post-meeting cleanup and item-by-item distribution. Supports Feishu native data + any meeting transcript, not tied to a single meeting platform.

✨ Features (MCP Tools)

Related MCP server: Feishu/Lark OpenAPI MCP

📦 Installation

1. Install Python dependencies

pip install mcp

2. Install and log in to lark-cli (Feishu command-line tool)

npm install -g lark-cli
lark-cli auth login

When logging in, scan the QR code with your own Feishu account to authorize, and check the following permission scopes (Scope):

vc:meeting.search:read
vc:meeting.meetingevent:read
vc:record:readonly
minutes:minutes.search:read
minutes:minutes.basic:read
minutes:minutes.artifacts:read
minutes:minutes.transcript:export
vc:note:read
docs:document.content:read

It's fine if you don't select everything at once — you can grant additional permissions later when prompted when using a specific feature.

🔌 Connect to your AI client

Claude Code (VS Code extension / CLI)

Add to mcpServers in settings.json (user or project level):

{
  "mcpServers": {
    "meeting-summary": {
      "command": "python",
      "args": ["D:/meeting-summary-mcp/server.py"]
    }
  }
}

Claude Desktop / other MCP-compatible clients

Point to the same in the client's MCP config: command: python, args: [absolute path to server.py].

🚀 Usage Examples

Scenario 1: Search today's / a specific day's meetings

Have AI call search_meetings("2026-08-21", "2026-08-21") to get the meeting list for that day.

Scenario 2: Turn a minutes/notes link into department-based to-dos

1. Give the meeting minutes link (.../minutes/obcnxxx) to AI

2. Tell AI the attendees' roles, for example:

   吴楠是北京产品经理，熊燕刚是运营总监，季子禾是直播间运营，
   刘晓曼是品牌运营，紫芙BB是品牌对接，王晨旭是直播间运营负责人

3. AI calls organize_todos_by_department, passing minute_token or link and roles_json, and gets:

# 会议待办 · 按部门分发

## 产品/技术部
- [ ] 积分问题排查：…（负责人：吴楠·北京产品经理）

## 直播间运营部
- [ ] 直播间GMV录入：…（负责人：王晨旭·直播间运营负责人）

Scenario 3: Read the transcript for deep analysis

Call fetch_transcript(minute_token) to get the full text, then have AI independently analyze based on the raw speech (more accurate than directly copying the AI summary).

Scenario 4: Any meeting such as Tencent Meeting (no Feishu dependency)

Hand the meeting transcript exported from Tencent Meeting (txt) directly to AI to read with import_meeting_text, then call organize_todos_from_text combined with the role config to organize to-dos by department:

张三 00:00:10
这个需求需要尽快跟进，明天给产品确认。
李四 00:00:20
培训文档我来输出，会后整理下发给运营。

→ organize_todos_from_text(text, roles_json, title) outputs department-grouped to-dos. WeChat meetings / voice-to-text transcriptions / any transcript work the same way.

🧩 roles_json Role Config Format

The roles_json parameter for organize_todos_by_department:

{
  "吴楠": {"role": "北京产品经理", "department": "产品/技术部"},
  "熊燕刚": {"role": "运营总监", "department": "运营总监"},
  "附子LC-直播运营": {"role": "直播间运营", "department": "直播间运营部"},
  "刘晓曼": {"role": "品牌运营", "department": "品牌运营/品牌对接"}
}

• Key: attendee name (display name in the meeting) or open_id (starting with ou_)

• role: job title; department: department (to-dos are grouped by this)

• To-dos that don't match any department are automatically placed under "To be confirmed / Needs manual assignment"

See examples/roles_example.json for an example

📄 Output Format (Standard Four-Section Template)

The standard template output by organize_todos_by_department / organize_todos_from_text:

# 会议纪要 · {会议主题}

## 一、会议信息      （主题 / 时间 / 参会人）
## 二、会议摘要      （一段话概括）
## 三、会议内容      （按议题分节）
## 四、待办事项 · 按部门分发
### 🛠️ 部门A
- [ ] 任务（负责人：xxx·岗位）
### 待确认 / 需人工认领

Template description and a real example see examples/meeting_template.md

📁 Project Structure

meeting-summary-mcp/
├── server.py            # MCP 入口（FastMCP + 8 个 tools）
├── feishu_client.py     # lark-cli 封装（搜索/妙记/纪要/逐字稿）
├── organizer.py         # 按部门待办整理逻辑
├── requirements.txt
├── README.md
├── examples/
│   ├── roles_example.json   # 岗位配置示例
│   └── output_example.md    # 生成的按部门待办示例
└── test/
    └── smoke_test.py        # 本机冒烟测试

🧪 Self-Test

cd meeting-summary-mcp
python test/smoke_test.py

Requires lark-cli to be logged in locally with the corresponding scopes authorized.

🔄 Sharing with Others

Simply package the entire meeting-summary-mcp directory (zip) and send it to the other person. They can follow the three steps in this README — "Install → Authorize → Connect" — to use it with their own AI client + their own Feishu account.

Note: This MCP accesses Feishu via lark-cli. Data is only read within your own account's authorized scope, and meeting content is never leaked across accounts.