sales-chat-quality
Click on "Install 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., "@sales-chat-qualityCollect today's sales chats and score them for quality."
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.
SalesChatQuality - 客服聊天质检自动化 Skill
用于解决售前客服会话依赖人工逐条抽检、回复时延计算效率低、质检结论缺少原文证据的问题。完成真实客服会话端到端闭环验证后,项目已交付前同事并在实际客服质检工作中使用,将会话筛选、详情读取、回复时延计算、语义质检、证据整理和报告输出中的大量步骤自动化,显著减少人工质检介入。
当前版本仍是小规模初版 MVP,不代表已形成大规模企业级批量生产系统。首轮以
1条真实会话验证采集、解析、计时、评价、证据和报告链路能够闭环;该数字仅是技术验证细节,不是项目使用范围。公开仓库已脱敏,不包含后台域名、账号、客户数据或真实截图。
解决方案
已登录 Chrome
│ CDP / read-only
▼
collect_chat.mjs ──► conversations.json + screenshots
│
▼
score_chats.py ────► response_seconds + rule result
│
▼
LLM + rubric ──────► evidence-linked quality review
│
▼
traceable report ──► JSON / PPTX浏览器自动化负责采集:进入售前查询、打开详情、等待加载、截图并解析角色/时间/正文。
确定性代码负责时间计算:阈值可配置,异常时间戳不会被当作正常结果。
LLM 负责语义评价:检查需求理解、准确性、费用说明、下一步引导、语气与问题覆盖,并要求引用原句。
MCP 负责工具化接入:以 stdio 暴露采集和评分入口,供 Codex 或其他 MCP 客户端编排。
Related MCP server: Salesloft MCP Demo Server
30 秒验证评分逻辑
无需登录任何后台即可运行脱敏样例:
python scripts/score_chats.py examples/conversations.sample.json --output examples/scored.sample.json --slow-threshold 10示例包含两次响应:13 秒会被标记为“回复慢”,10 秒被标记为“正常”。运行后还会生成 examples/evaluation_prompt.txt,用于约束 LLM 的证据化评价。
运行测试:
python -m unittest discover -s tests -v使用真实后台采集
安装依赖:
npm install
python -m pip install -r requirements.txt用隔离用户目录启动 Chrome,并在该窗口自行登录后台:
& "$env:ProgramFiles\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222 --user-data-dir="$env:TEMP\codex-sales-chat-profile"采集只读证据:
node scripts/collect_chat.mjs --cdp-url http://127.0.0.1:9222 --max-conversations 20 --output "$env:USERPROFILE\Desktop\售前聊天质检"采集器针对一个真实后台的页面结构开发。迁移到其他系统时,需要根据菜单文案、DOM、分页和详情弹窗做局部适配;它不宣称零修改兼容所有客服后台。
MCP 接入
{
"mcpServers": {
"sales-chat-quality": {
"command": "python",
"args": ["C:/path/to/SalesChatQuality/scripts/mcp_server.py"]
}
}
}工具:
collect_sales_chats:连接已登录的 Chrome CDP 会话并输出截图和结构化 JSON。score_sales_chats:计算响应时间,返回评分文件、评价提示词和异常数量。
仓库结构
SalesChatQuality/
├── SKILL.md # Codex Skill 工作流与安全边界
├── agents/openai.yaml # Skill UI 元数据
├── references/evaluation-rubric.md # 语义质检口径
├── scripts/collect_chat.mjs # Playwright/CDP 采集器
├── scripts/score_chats.py # 确定性响应时间计算
├── scripts/mcp_server.py # stdio MCP Server
├── examples/ # 脱敏输入样例
└── tests/ # 评分边界测试实际使用与 MVP 验证结果
初版 MVP 已交付前同事并在实际客服质检工作中使用,将会话读取、计时、语义质检、证据整理与报告输出中的大量步骤自动化。
受首版交付时间限制,先选取
1条真实会话完成端到端回归;该数字只用于说明技术链路验证,不代表使用范围。该会话解析出
3条买家消息和4条售前消息;初版规则得到13秒慢回复和10秒正常回复,阈值可按业务 SOP 调整。当前使用规模较小,不将项目外推为批量覆盖、大规模企业级生产系统或生产级准确率。
公开仓库不提供真实截图或客户数据。运行时生成的 JSON、截图与报告默认被
.gitignore排除。
后续迭代
增加批量会话采集、定时调度、失败重试与断点续跑。
将质检 SOP、阈值和评价维度版本化,支持按业务线动态配置。
建立人工复核样本集,持续评估规则与 LLM 评价的一致性和准确率。
补充后台 DOM 适配层、异常监控和报告看板,降低页面变化带来的维护成本。
安全边界
不读取或保存密码、Cookie、浏览器配置或验证码。
只执行查询、打开详情和截图,不发送消息或修改后台数据。
页面结构变化时停止操作并保存调试证据,不猜测点击。
对外分享前必须移除后台域名、手机号、IP、账号和客服真实姓名。
完整操作流程与失败处理见 SKILL.md,评价维度见 references/evaluation-rubric.md。
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
CX ops: read conversations, calls and QA evaluations from Zendesk, Freshdesk, Five9 and more.
Sentiment, toxicity, entity extraction, PII, translation, summary, QA, fraud scoring, safety audit.
Coach your communication from your own recorded conversations: talk-time, questions, rapport, more.
Sales intelligence for B2B SMEs — lead scoring, ICP fit, CRM enrichment & writeback.
Related MCP Servers
- AlicenseNot gradedqualityNot gradedmaintenanceEnables intelligent conversation management with 4 AI agents that provide semantic analysis, pattern discovery, automatic documentation, and relationship mapping. Logs and analyzes Claude conversations with 70% token optimization and multi-language support.2
- FlicenseAqualityDmaintenanceEnables AI-powered analysis of sales call transcripts through natural language queries, allowing users to search conversations, identify objections, and extract insights from customer calls without manually reviewing recordings.3
- AlicenseNot gradedqualityCmaintenanceEnables AI-driven data quality operations such as validation, enrichment, and deduplication for customer data fields including email, address, phone, and LinkedIn profiles through natural language interactions.2MIT
- AlicenseAqualityAmaintenanceEnables AI-powered interaction with Salesforce, providing opportunity intelligence, engagement analysis, business case generation, and standard CRUD operations through natural language.18605MIT
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/Yuan1cMax/SalesChatQuality'
If you have feedback or need assistance with the MCP directory API, please join our Discord server