qa-workflow
This server facilitates an automated QA bug-fixing workflow where Claude Code waits for bug reports, fixes them, and loops continuously until explicitly stopped.
MCP Tools:
wait_for_bug_report— Blocks and waits for the next bug report from the pending queue (submitted via HTTP API); returns the full bug object (id,sourceLocation,description,element, etc.) so Claude Code can precisely locate and fix the issuecomplete_bug— Marks a bug as fixed by archiving it to the completed list; accepts the original bug object and an optionalfixSummary(e.g.,file:line — description of change)get_queue_status— Returns the current state of both pending and completed bug queues
HTTP API (for external systems/testers):
POST /api/bug-report— Submit a new bug report (source location, description, element info, etc.)GET /api/status— View overall queue statusGET /api/pending— List pending bugsGET /api/completed— List completed/archived bugs
All bug data is persisted to disk, and the workflow is designed as a continuous loop: wait → fix → archive → repeat, exiting only on explicit user command.
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., "@qa-workflow进入提测"
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.
agentation-qa-skill
QA 提测工作流 MCP Server — 配合 agentation-zero 使用,让 Claude Code 自动等待测试反馈、精准定位代码、循环修复 bug。
工作原理
测试人员在页面圈选元素 (agentation-zero Toolbar)
↓
标注数据(含文件路径 + 行号)通过 POST /api/bug-report 推送给 agentation-qa-skill
↓
写入 data/pending/bugs.json,唤醒 MCP 进程
↓
Claude Code 读取 sourceLocation 精准跳转 → 修复 → complete_bug 归档
↓
complete_bug 返回后立即再次 wait_for_bug_report,持续等待下一条 bug进入提测 后,Claude Code 会进入不可中断的修复循环:等待 bug、修复、归档、再次等待。只有用户明确发送 结束提测 才会退出循环。
双进程架构
进程 | 入口 | 职责 |
| Claude Code 通过 | MCP 工具暴露(stdio,stdout 完全干净) |
|
| HTTP API,接收测试人员提交的 bug |
两个进程共享 data/ 目录下的 JSON 文件通信,互不干扰。
Related MCP server: mcp-flow
快速开始
1. 安装 agentation-qa-skill
git clone <this-repo> agentation-qa-skill
cd agentation-qa-skill
npm install2. 在业务项目中安装 agentation-zero
npm install agentation-zero --save-dev3. 配置 vite.config.ts
import { defineConfig } from 'vite'
import react from '@vitejs/plugin-react'
import agentationLocator from 'agentation-zero/plugins/locator'
export default defineConfig({
plugins: [
agentationLocator(),
react(),
],
})4. 挂载 Toolbar
import { Agentation } from 'agentation-zero'
import 'agentation-zero/dist/style.css'
function App() {
return (
<>
<Agentation />
</>
)
}5. 配置 Claude Code
安装 skill:
mkdir -p your-project/.claude/skills/agentation-qa-skill
cp /path/to/agentation-qa-skill/SKILL.md \
your-project/.claude/skills/agentation-qa-skill/SKILL.md配置 MCP 连接,在业务项目根目录新建 .mcp.json:
{
"mcpServers": {
"agentation-qa-mcp-server": {
"command": "node",
"args": ["src/mcp.js"],
"cwd": "/absolute/path/to/agentation-qa-skill"
}
}
}
cwd必须是 agentation-qa-skill 的绝对路径,这是 MCP 工具无法暴露最常见的原因。
使用流程
# 终端 1:启动 HTTP API(接收 bug 提交)
cd agentation-qa-skill && npm start
# 终端 2:启动业务项目
cd your-project && npm run dev
# 在 Claude Code 中发送:
进入提测Claude Code 自动连接 agentation-qa-mcp-server,调用 wait_for_bug_report 阻塞等待。
测试人员通过 agentation-zero Toolbar 圈选元素提交后,Claude Code 自动唤醒、修复、调用 complete_bug 归档;complete_bug 返回后的下一个动作必须是再次调用 wait_for_bug_report,不能停下来等待用户确认,也不能输出等待提示。
结束时发送:结束提测
提测循环规则
进入提测后立即调用wait_for_bug_reportwait_for_bug_report阻塞是正常行为,不代表卡死收到 bug 后按
sourceLocation和element定位,只修复description描述的问题修复完成或无法修复时,调用
complete_bug(bug, fixSummary)complete_bug返回后必须立即再次调用wait_for_bug_report只有用户发送
结束提测才能调用get_queue_status并退出循环
排查 MCP 工具无法暴露
按以下顺序检查:
1. cwd 路径是否正确
# 确认路径存在且含 src/mcp.js
ls /absolute/path/to/agentation-qa-skill/src/mcp.js2. 依赖是否已安装
cd /absolute/path/to/agentation-qa-skill && npm install3. .mcp.json 是否在业务项目根目录
ls your-project/.mcp.json4. 手动验证 mcp.js 能否正常启动
cd agentation-qa-skill && node src/mcp.js
# 应该无任何 stdout 输出,只挂起等待
# 有任何 stdout 输出都会破坏 stdio 通信HTTP API
方法 | 路径 | 说明 |
POST |
| 提交 bug(任意 JSON,自动生成 id + createdAt) |
GET |
| 队列状态 |
GET |
| 待处理列表 |
GET |
| 已修复列表 |
手动提交示例:
curl -X POST http://localhost:4299/api/bug-report \
-H "Content-Type: application/json" \
-d '{
"sourceLocation": "src/pages/merchant/hotel/list/index.tsx:226:14",
"description": "修改「重置」文案为「清空」",
"element": "\"重 置\""
}'项目结构
agentation-qa-skill/ ← 独立运行,不放入业务项目
├── SKILL.md ← 复制到 your-project/.claude/skills/agentation-qa-skill/
├── .mcp.json ← 模板,复制到 your-project/ 并修改 cwd
├── package.json
├── src/
│ ├── mcp.js ← MCP Server(纯 stdio,stdout 零输出)
│ ├── http.js ← HTTP API Server(npm start 启动)
│ └── store.js ← JSON 文件读写 + 串行写锁 + 唤醒队列
└── data/
├── pending/bugs.json ← 待修复队列(FIFO)
└── completed/bugs.json ← 已修复归档
your-project/ ← 业务项目
├── .claude/
│ └── skills/
│ └── agentation-qa-skill/
│ └── SKILL.md ← 从 agentation-qa-skill/SKILL.md 复制
├── .mcp.json ← cwd 指向 agentation-qa-skill 绝对路径
└── ...常见问题
Q:MCP 工具没有暴露给 Claude Code?
99% 是 .mcp.json 的 cwd 路径写错了,必须是绝对路径。参考上方「排查」章节。
Q:处理完 bug 后 Claude Code 自动退出了?
确认 .claude/skills/agentation-qa-skill/SKILL.md 已正确放置,并且内容包含强制循环规则:complete_bug 返回后必须立即再次调用 wait_for_bug_report。
Q:为什么 Claude Code 看起来一直挂起?
这是 wait_for_bug_report 在阻塞等待新 bug,属于正常状态。不要改用轮询,也不要用其他工具替代它。
Q:多个测试人员同时提交会冲突吗? 不会。所有写操作通过串行调度器排队,按提交顺序进入 FIFO 队列,Claude Code 逐条处理。
Q:HTTP 服务重启后 pending 里的 bug 会丢失吗?
不会。数据持久化在 data/pending/bugs.json,重启后自动恢复。
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 Servers
- AlicenseAqualityBmaintenanceMCP server that spawns autonomous Claude Code agents in GitHub repos, enabling task delegation with persistent state, multi-step workflows, and job monitoring.471632Apache 2.0
- FlicenseAqualityCmaintenanceA safe, local MCP server that lets Claude drive a controlled software-development loop (inspect, read, plan, patch, apply, check, analyze, fix, summarize) on a project, using deterministic tools and real diffs/test runs.101
- AlicenseAqualityBmaintenanceAn MCP server that turns Grok CLI into the execution agent for Claude Code, implementing an orchestrated execute-verify-autofix loop for autonomous development tasks.7MIT
- Alicense-qualityAmaintenanceMCP server that watches issue trackers and comments for 'Dear Claude' and spawns local Claude Code instances to execute tasks, posting results back to the original platform.69MIT
Related MCP Connectors
Voice-powered bug reporting with 13 MCP tools. Record bugs by talking; let AI find and fix them.
Official MCP server for Qase — manage test cases, runs, suites, defects via AI tools.
Augments MCP Server - A comprehensive framework documentation provider for Claude Code
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/WtecHtec/agentation-qa-skill'
If you have feedback or need assistance with the MCP directory API, please join our Discord server