GitLab MCP Server

Instructions

Implementation Reference

• src/gitlab_mcp/gitlab_client.py:297-312 (handler)

  The actual implementation of the list_pipelines method, which interacts with the GitLab API.

  def list_pipelines(
      self,
      project_id: int | str,
      status: str = None,
      ref: str = None,
      page: int = 1,
      per_page: int = 20,
  ) -> list[dict]:
      """GET /projects/:id/pipelines"""
      pid = self._resolve_project_id(project_id)
      params = {"page": page, "per_page": per_page}
      if status:
          params["status"] = status
      if ref:
          params["ref"] = ref
      return self._get_json(f"/projects/{pid}/pipelines", params=params)

• src/gitlab_mcp/server.py:505-535 (registration)

  The MCP tool registration and wrapper function for list_pipelines.

  @mcp.tool()
  def list_pipelines(project_id: int | str, status: str = None,
                     ref: str = None, page: int = 1, per_page: int = 20) -> str:
      """列出專案的 CI/CD Pipelines
  
      Args:
          project_id: 專案 ID 或路徑
          status: 狀態篩選（running, pending, success, failed, canceled, skipped）
          ref: 分支或標籤名稱篩選
          page: 頁碼
          per_page: 每頁筆數
      """
      try:
          if status:
              v = GitLabValidator.validate_pipeline_status(status)
              if not v.is_valid:
                  return "\n".join(v.errors)
  
          client = get_client()
          pipelines = client.list_pipelines(project_id, status=status, ref=ref,
                                            page=page, per_page=per_page)
          if not pipelines:
              return "找不到符合條件的 Pipeline"
  
          status_emoji = {
              "success": "✅", "failed": "❌", "running": "🔄",
              "pending": "⏳", "canceled": "⛔", "skipped": "⏭️",
              "created": "🆕", "manual": "👋"
          }
          lines = [f"找到 {len(pipelines)} 個 Pipeline：\n"]
          for p in pipelines: