Alibaba Cloud Yunxiao MCP Server (alibabacloud-devops-mcp-server) is an open-source DevOps tool provided by Alibaba Cloud that gives AI assistants the ability to interact with the Yunxiao platform. It encapsulates a set of tools for calling Yunxiao APIs, enabling AI assistants to read work item content in projects, automatically write code after understanding requirements, and submit code merge requests. Enterprise R&D teams can use it to assist with code review, optimize task management, and reduce repetitive operations, so they can focus more on innovation and product delivery.
[Main Capabilities] This service provides the following toolsets, each covering a different domain:
- organization-management: Organization management tools, including organization list, organization details, department information, organization roles, member information, etc.
- code-management: Code repository management tools, including repository management, branch management, merge request management, file tree, etc. The specific tools include create_branch, delete_branch, get_branch, list_branches, create_file, delete_file, get_file_blobs, list_files, update_file, create_change_request, update_change_request, create_change_request_comment, get_change_request, list_change_request_patch_sets, list_change_request, list_change_request_comments, get_compare, get_repository, list_repositories, list_commits, get_commit, create_commit_comment, etc.
- project-management: Project management tools, including projects, work items, work item fields, work item comments, effort tracking, etc. The specific tools include get_project, search_projects, get_sprint, list_sprints, create_sprint, update_sprint, search_programs, list_program_versions, list_versions, create_version, update_version, delete_version, get_work_item, update_work_item, search_workitems, get_work_item_types, create_work_item, list_all_work_item_types, list_work_item_types, get_work_item_type, list_work_item_relation_work_item_types, list_workitem_relation_records, create_workitem_relation_record, delete_workitem_relation_record, get_work_item_type_field_config, get_work_item_workflow, list_work_item_comments, create_work_item_comment, list_workitem_attachments, get_workitem_file, create_workitem_attachment, list_workitem_activities, as well as effort-related tools (list_current_user_effort_records, list_effort_records, create_effort_record, list_estimated_efforts, create_estimated_effort, update_effort_record, update_estimated_effort), etc.
- pipeline-management: Pipeline management tools, including pipeline list, pipeline configuration, resource management, tag management, deployment management, etc. The specific tools include get_pipeline, list_pipelines, smart_list_pipelines, generate_pipeline_yaml, create_pipeline_from_description, update_pipeline, create_pipeline_run, get_latest_pipeline_run, get_pipeline_run, list_pipeline_runs, list_pipeline_jobs_by_category, list_pipeline_job_historys, execute_pipeline_job_run, get_pipeline_job_run_log, list_service_connections, create_resource_member, delete_resource_member, list_resource_members, update_resource_member, update_resource_owner, create_tag, create_tag_group, list_tag_groups, delete_tag_group, update_tag_group, get_tag_group, delete_tag, update_tag, as well as VM deployment-related tools (stop_vm_deploy_order, skip_vm_deploy_machine, retry_vm_deploy_machine, resume_vm_deploy_order, get_vm_deploy_order, get_vm_deploy_machine_log), etc.
- application-delivery: Application delivery tools, including deployment order management, application management, application tags, variable group management, etc. The specific tools include create_change_order, list_change_order_versions, get_change_order, list_change_order_job_logs, find_task_operation_log, execute_job_action, list_change_orders_by_origin, as well as change request-related tools (create_appstack_change_request, get_appstack_change_request_audit_items, list_appstack_change_request_executions, list_appstack_change_request_work_items, cancel_appstack_change_request, close_appstack_change_request), and application orchestration-related tools (list_applications, get_application, create_application, update_application, get_latest_orchestration, list_app_orchestration, create_app_orchestration, delete_app_orchestration, get_app_orchestration, update_app_orchestration), etc.
- packages-management: Artifact repository management tools (artifact repositories, artifact lists, etc.).
- test-management: Test management tools (test case management, test case directories, test plans, test results, etc.).
[Usage]
- Version support: Supports two deployment modes: Yunxiao Central Station and Region Edition. Central Station uses https://openapi-rdc.aliyuncs.com as the API domain; Region Edition uses an organization-specific domain, such as https://your-org.devops.aliyuncs.com。工具根据配置的 The API base URL automatically determines the deployment mode: if the URL contains openapi-rdc.aliyuncs.com, it is Central Station mode; otherwise, it is Region Edition mode. When using Region Edition, you need to set the environment variable YUNXIAO_API_BASE_URL.
- Prerequisites: Node version ≥ 18.0.0; an Alibaba Cloud Yunxiao Personal Access Token is required, with read and write permissions granted for all APIs under organization management, project collaboration, code management, pipeline management, artifact repository management, application delivery, and test management.
- Official hosted MCP service (no installation required): Yunxiao provides an official hosted MCP endpoint, requiring no local installation or Docker. Central Station endpoint:https://openapi-rdc.aliyuncs.com/ai/mcp;Region Edition endpoint: organization Region domain + /ai/mcp, for example https://<your-org>.devops.aliyuncs.com/ai/mcp. The transport method is Streamable HTTP (stateless); authentication uses Authorization: Bearer <YOUR_TOKEN> or the request header X-Yunxiao-Token: <YOUR_TOKEN>. You can request only the required toolsets via ?toolsets=code-management,project-management or the request header X-Devops-Toolsets to reduce context usage.
- Self-hosted Stdio mode (simplest): Suitable for most MCP clients (such as Cursor, Claude Desktop, iFlow, etc.), no Docker required, run via npx. Add the configuration in the MCP client configuration file, replace <YOUR_TOKEN> with the Yunxiao access token, and the -y flag automatically confirms the installation. By default, it calls the Yunxiao Central Station OpenAPI (https://openapi-rdc.aliyuncs.com);若为 Region edition requires an additional environment variable YUNXIAO_API_BASE_URL.
- Docker (optional): You can build the image yourself with docker build -t alibabacloud/alibabacloud-devops-mcp-server ., and set the transport mode at startup via MCP_TRANSPORT and PORT. Optional Streamable HTTP environment variables include MCP_STREAMABLE_PATH (default /mcp), MCP_HTTP_HOST (default 0.0.0.0), and MCP_ALLOWED_HOSTS (comma-separated).
- Transport modes: Streamable HTTP is the remote transport method recommended by the MCP specification, while SSE is the traditional method; both can be used simultaneously during the migration period. Client URL: for SSE it is http://localhost:3000/sse,Streamable HTTP is http://localhost:3000/mcp。
- Toolset selection: The service supports toolsets, and you can enable only the required tools to reduce the number of tools presented to the AI assistant and improve performance. You can specify the toolset via command-line arguments or environment variables; if not specified, all tools are enabled by default.
[Use cases] Aimed at enterprise R&D teams, it is used to assist AI assistants in performing code reviews, optimizing task management, and reducing repetitive operations on the Yunxiao platform. Specifically, it can read project work item content, automatically write code after understanding requirements and submit merge requests, and can also manage code repositories, projects, pipelines, artifact repositories, application delivery, and testing, etc.
[Limitations] Requires Node version ≥ 18.0.0; requires a Yunxiao personal access token with the corresponding read and write permissions; using the Region edition requires additional configuration of environment variables or a domain name. Other information such as fees and pricing is not currently provided.