MCP_HCL
Provides tools for interacting with H3C Cloud Lab (HCL) project files: listing and read-only inspecting existing .net projects (devices, coordinates, links), and staging additive changes by generating independent copies of a project that add a device (copied from an existing same-model device in the project) or add a dual-end link between two free ports. Callers then open the resulting .net file in HCL to validate the topology; the server does not modify source projects in place or control the live HCL canvas.
Click on "Deploy 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., "@MCP_HCL add a new router to project Lab1 and link it to switch1"
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.
MCP_HCL
与 Agent 客户端无关的 HCL MCP Server,当前版本 0.5.0。本地主分支整合了已有分支历史和最新已验证的软件成果;完整 MCP-Packet-Tracer 62 工具/5 资源等价仍为 Incomplete。
接手读 HANDOFF.md,完整索引见 docs/README.md。当前发现面为 55 工具/6 资源;发现、软件合同、原生配置与真实流量分别记账,full_live_gate_passed=false。
能力与限制
离线工程检查、阶段副本、编辑、严格规划与配置/操作清单生成,支持哈希绑定的导出和重载。
原生配置服务包含 ACL/VLAN/NAT、接口调优、STP/端口安全与 NetStream;默认写入关闭,必须绑定实际设备、备份、范围回读和恢复入口。
差异、健康、地址观察和安全审计保留未知值及模拟/原生来源;不把缺字段、阶段库存或保存文件解释为完整实时拓扑。
GUI 保护验证身份、当前账号目录、监听者及工程回执;不能证明厂商运行时环境或清理边界。
上游工具库存仍为 12 implemented / 47 planned / 3 unverified。原生完整设备/线缆库存、若干 IOS→Comware 行为映射、命名 exporter/monitor 与安全 Phone 关联等仍缺。
2026-10-11 具名报告记录一次单设备隔离生命周期和账号清理;候选独占 CLI 被后台 MultiCC 保护拒绝,配置与网络验收未完成。证据位置与后续门槛见 当前交接 和 测试指南。
Related MCP server: workrelay
安装与运行
需要 Python 3.11+;原生功能需要合法安装的 Windows HCL。使用当前工作区的独立环境:
python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -e '.[dev]'
$env:MCP_HCL_PROJECT_ROOTS = 'C:\Users\YOUR_NAME\HCL\Projects'
$env:MCP_HCL_OUTPUT_ROOT = (Join-Path (Get-Location) 'work\staged')
$env:MCP_HCL_INSTALL_ROOT = 'D:\HCL'
$env:MCP_HCL_ALLOW_LIVE_WRITE = '0'
.\.venv\Scripts\python.exe -m mcp_hcl.serverstdio 服务等待协议输入;在 MCP 客户端配置相同解释器、工作区和环境。阶段工具仅写隔离输出副本。原生写入须按 统一测试计划 处理身份、资产和恢复前提。
开发与验证
.\.venv\Scripts\python.exe scripts\check_all.py该入口运行 Ruff 和全部软件检查,强制真实写开关为 0,并为每轮保存独立回执。历史成功与失败不覆盖。开发规范和模块职责见 development.md。原始检查结果定位见 HANDOFF。
参数见 离线 API、严格合同 v2 和 Schema。日期化报告、重复交接与旧脚本快照已移出维护文档,保留 Git 历史和仓库外备份。
Available Tools
5 toolshcl_capabilitiesC
Describe verified capabilities and the activation boundary.
Inputs: no named parameters. Returns: object
hcl_inspect_projectB
Read devices and links from one saved HCL project without changing it.
Inputs: project_path (required). Returns: object
hcl_list_projectsA
List HCL projects immediately under configured project roots.
Inputs: no named parameters. Returns: object
hcl_stage_add_deviceC
Copy a project, add a device cloned from a known model, and optionally link it.
Inputs: x (required), y (required), new_name (required), stage_name (required), +5 more. Returns: object
hcl_stage_linkC
Copy a project and add a reciprocal link between two free ports.
Inputs: a (required), b (required), a_port (required), b_port (required), +2 more. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 5 tools
Each tool has a clearly distinct purpose: capabilities (meta), list_projects (enumerate), inspect_project (read), stage_add_device (add device), and stage_link (add link). The optional linking in stage_add_device could be confused with stage_link, but the descriptions clarify the primary actions and when to use each.
All tools use the hcl_ prefix and snake_case, with operational tools following a verb_noun pattern (list_projects, inspect_project, stage_add_device, stage_link). hcl_capabilities is a noun phrase and breaks the verb-based convention, but only as a minor deviation.
Five tools is well-scoped for a server focused on HCL project inspection and staged modifications. Each tool has a clear role, and there is no redundant or excessive tooling.
The surface covers listing, inspection, and additive staging (add device, add link), but lacks update, removal, or project creation/deletion operations. For a tool that manipulates HCL projects, these are notable gaps in lifecycle coverage.
Related MCP Connectors
Read project context and manage tasks in MCPBinder, with access you control.
Project management shared by people and AI agents, with persistent project state through MCP.
Start and manage Reason cloud coding agents, projects and sessions from any MCP client.
Verified component catalog over remote MCP: agents pick ids and get a project that builds.
Related MCP Servers
- AlicenseBqualityBmaintenanceEnables MCP-compatible agents to perform read-only inspection of OpenPLC Editor projects, including retrieving project structure and listing programs, function blocks, and functions.3Apache 2.0
- AlicenseNot gradedqualityCmaintenanceEnables external agents to inspect and interact with a local single-project workbench via MCP, providing read-only tools for state, audit, resume, readiness, comparison, impact preview, and repair candidates, while proposals are saved for manual confirmation in the workbench.MIT
- AlicenseCqualityBmaintenanceEnables users to plan and stage H3C Cloud Lab topologies offline, inspect a running lab read-only through its telnet bridge, and change device configuration only via gated, receipt-backed operations that are disabled by default. Ships a DeepSeek Harness plugin bundle and also works with any stdio MCP client.5517 npm3MIT
- AlicenseNot gradedqualityCmaintenanceEnables MCP-compatible clients to plan, validate, generate, and deploy complete network topologies from natural-language instructions, including live deployment to a running network simulator and export to configuration or script files.MIT