Polarion MCP

mcp极化子

CLI-first Polarion ALM客户端，为AI助手（Cline、Claude Code、Copilot Studio）提供MCP服务器支持，并为OpenAI GPT操作和其他HTTP消费者提供REST API。

安装

git clone https://github.com/mmerah/PolarionMcp.git
cd PolarionMcp
python -m venv .venv && source .venv/bin/activate
pip install -e .

创建一个 .env 文件：

POLARION_URL=https://your-polarion-instance.com/polarion
POLARION_USER=your-username
POLARION_TOKEN=your-personal-access-token

CLI使用情况

# Connection
mcp-polarion health

# Projects
mcp-polarion projects                       # list all configured aliases
mcp-polarion project -p myproject           # name & description
mcp-polarion project-types -p myproject     # configured work item types + fields
mcp-polarion named-queries -p myproject     # named queries defined in config
mcp-polarion discover-types -p myproject    # sample the project to find types
mcp-polarion discover-types -p myproject --limit 500

# Work items
mcp-polarion get ADC-1234                   # project inferred from ID prefix
mcp-polarion get ADC-1234 -p myproject
mcp-polarion search "type:defect AND status:open" -p myproject
mcp-polarion search "query:open_bugs" -p myproject  # named query
mcp-polarion search "type:defect" -p myproject --fields "id,title,status"

# Tests & documents
mcp-polarion test-runs -p myproject
mcp-polarion test-run TR-42 -p myproject
mcp-polarion documents -p myproject
mcp-polarion test-specs QA/TestSpecs -p myproject

# Plans (plan projects only)
mcp-polarion plans -p releases
mcp-polarion plan R2024.4 -p releases
mcp-polarion plan-workitems R2024.4 -p releases
mcp-polarion search-plans "templateId:release" -p releases

# Servers
mcp-polarion serve                          # HTTP works for all clients
mcp-polarion serve --mode stdio             # stdio transport for local .mcp.json
mcp-polarion serve --port 9000 --log-level DEBUG

# Utilities. Output goes to generated/
mcp-polarion docgen                         # generated/agent_instructions*.md
mcp-polarion generate-openapi               # generated/openapi.yaml
mcp-polarion import-custom-fields --path local/custom-fields/myproject

便利脚本 run_server.sh 处理首次venv设置和 .env 通话前检查 mcp-polarion serve.

配置

可选的本地配置解锁了项目别名、命名查询、自定义字段映射和计划项目支持。被追踪的例子存在于 local/。复制它以开始：

cp local/polarion_config.example.yaml local/polarion_config.yaml

projects:
  myproject:                          # alias (use this everywhere)
    id: ACTUAL_PROJECT_ID             # real Polarion project ID
    work_item_types:
      - systemRequirement
      - defect
    custom_fields:
      defect: [severity, foundIn]
      systemRequirement: [acceptanceCriteria, riskRelevance]
    default_queries:
      open_bugs: "type:defect AND status:open"
      my_items: "assignee.id:$current_user"

  releases:
    id: RELEASES_PROJECT
    is_plan: true                     # enables plan-specific tools

display_fields: [id, title, type, status, assignee]

编辑配置后，重新生成代理指令和OpenAPI规范：

mcp-polarion docgen                         # generated/agent_instructions*.md
mcp-polarion generate-openapi               # generated/openapi.yaml

对于自定义字段导入，请将下载的XML导出保持在本地且不被跟踪：

local/custom-fields/
/
  requirement-custom-fields.xml
  defect-custom-fields.xml

通常，这些custom-fields.yml位于Polarion项目管理内容中。它们可以从那里下载。然后使用以下方式导入它们：

mcp-polarion import-custom-fields --path local/custom-fields/myproject

看 docs/XML_PARSER.md 完整的导入工作流程。

可用工具

看 docs/WORKFLOW_EXAMPLES.md 了解工具使用模式。

将军

工作项 *（常规项目）*

测试和文件

计划 *（仅计划项目）*

输入: project_alias 接受两个别名(myproject)和真实ID(ACTUAL_PROJECT_ID).\ 输出：人类可读的字符串。错误以开头 ❌.

MCP服务器集成

一个单一的 mcp-polarion serve 命令为所有HTTP客户端提供服务。服务器使用无状态请求（每个POST一个会话）和JSON响应，从而实现并行工具调用，并同样适用于Cline、Claude Code、Copilot Studio和GPT Actions（通过REST API）。

Cline/Claude代码（HTTP）

mcp-polarion serve          # http://0.0.0.0:8000/mcp

添加到Cline的MCP设置中：

{
  "mcpServers": {
    "polarion": { "url": "http://localhost:8000/mcp" }
  }
}

克劳德代码/ .mcp.json （stdio）

{
  "mcpServers": {
    "polarion": {
      "command": "mcp-polarion",
      "args": ["serve", "--mode", "stdio"]
    }
  }
}

微软复制品工作室

mcp-polarion serve

通过HTTPS公开（例如Azure Dev Tunnel），然后按照 MCP入职向导: 添加工具→ 新工具→ 模型上下文协议.使用 https:///mcp 作为服务器URL。

REST API/GPT操作（OpenAI）

mcp-polarion serve

REST端点可在 /actions/ 在MCP端点旁边。首先生成OpenAPI规范和代理指令：

mcp-polarion generate-openapi               # generated/openapi.yaml
mcp-polarion docgen                         # generated/agent_instructions*.md

正在运行的服务器也满足以下规范 /openapi.json 和 /openapi.yaml.

curl http://localhost:8000/actions/projects
curl http://localhost:8000/openapi.json | jq '.paths | keys'

使用 generated/agent_instructions.md 作为知识档案和 generated/agent_instructions_simple.md 系统提示您自定义GPT。

项目结构

polarion_mcp/
  core/           Polarion domain logic: client, config, settings, formatters
  config/         Config discovery and XML import utilities
  mcp/            MCP layer: tool definitions, middleware, server, stdio
  rest_api/       REST API routes and OpenAPI spec generation
  docgen/         Agent instruction doc generator
  cli/            CLI entry point (mcp-polarion)
docs/             Supporting guides and workflow references
local/            Workspace-local config and XML exports
generated/        Generated files (openapi.yaml, agent_instructions)
tests/
run_server.sh              First-time bootstrapper (venv + deps + serve)

壳牌完井

子命令和标志的基本选项卡完成可通过以下方式获得 argcomplete.重新安装软件包后，在shell中启用它：

pip install -e .
eval "$(register-python-argcomplete mcp-polarion)"

对于bash，将该行添加到 ~/.bashrc.对于zsh，首先启用bash补全兼容性，然后注册相同的命令。

发展

pytest
black polarion_mcp tests
isort polarion_mcp tests
mypy polarion_mcp

许可证

MIT。看 许可证.