huawei-lts-skill
Enables interaction with Huawei Cloud Log Tank Service (LTS) through natural language, supporting log group listing, log search, and other LTS operations.
README
Huawei LTS Skill + MCP
A portable Agent Skill and lightweight stdio MCP server for Huawei Cloud Log Tank Service (LTS). It supports Cursor, Claude Code, OpenCode, and Codex without cloning the full upstream Huawei Cloud MCP repository.
What is included
- Root
SKILL.mdwith progressive references and safe LTS workflows. - A Python MCP server exposing 92 LTS operations backed by the official Huawei Cloud Python SDK.
- Automatic macOS/Linux installers for four coding agents.
- Locked runtime dependencies and tests that require no cloud credentials.
The runtime dependencies are limited to MCP Python SDK v1,
huaweicloudsdkcore, and huaweicloudsdklts, plus their required transitive
packages.
Requirements
- macOS or Linux
- Python 3.10 or newer
uv- At least one supported client
- Huawei Cloud AK/SK with permissions appropriate to the operations you use
Install
git clone https://github.com/<owner>/huawei-lts-skill.git
cd huawei-lts-skill
./install.sh
Install selected clients only:
./install.sh --platforms cursor,claude
./install.sh --platforms opencode,codex
The installer deploys the runtime to ~/.local/share/huawei-lts-skill, creates
an isolated .venv, installs the Skill, backs up existing client configs, and
merges the huawei-lts MCP entry. Override the deployment location with
HUAWEI_LTS_INSTALL_DIR=~/some/subdirectory.
Configure credentials
After installation, replace these placeholders in every selected client's MCP configuration:
HUAWEI_ACCESS_KEY=REPLACE_WITH_YOUR_AK
HUAWEI_SECRET_KEY=REPLACE_WITH_YOUR_SK
HUAWEI_REGION=REPLACE_WITH_YOUR_REGION
HUAWEI_PROJECT_ID=REPLACE_WITH_YOUR_PROJECT_ID
Configuration locations:
| Client | Skill directory | MCP configuration |
|---|---|---|
| Cursor | ~/.cursor/skills/huawei-lts |
~/.cursor/mcp.json |
| Claude Code | ~/.claude/skills/huawei-lts |
~/.claude.json |
| OpenCode | ~/.config/opencode/skills/huawei-lts |
~/.config/opencode/opencode.json |
| Codex | ~/.codex/skills/huawei-lts |
~/.codex/config.toml |
Do not commit client configurations containing real credentials. Restart the client after editing its configuration.
Use
Use huawei-lts to list log groups in {region}, project_id={project_id}.
Search {log_group_id}/{log_stream_id} for {keyword} during the last hour.
Give me a concise conclusion and redacted examples.
The Skill defaults to read-only inspection. Update, Delete, enable/disable, and batch Create operations require an explicit impact summary and user confirmation.
Upgrade and uninstall
Pull or download the new repository version and rerun ./install.sh. Installation
is idempotent and backs up affected client configs before updating them.
./uninstall.sh
./uninstall.sh --platforms cursor,codex
Uninstall removes only this Skill, its MCP entries, and its deployed runtime. It does not touch Huawei Cloud resources or unrelated client configuration.
Development
uv sync --extra dev
uv run pytest
uv run python scripts/generate_tools.py --check
The checked-in tool manifest matches the current 92-tool MCP surface. When the Huawei LTS SDK changes, update the manifest deliberately and run the catalog validation before release.
Troubleshooting
uv is required: install uv and rerun the installer.- MCP does not start: confirm the configured command exists under the deployed
.venv/bin/directory and restart the client. - Missing credential/scope error: replace all four
HUAWEI_*placeholders. - Empty resource list: verify that region and project ID belong together.
- Permission error: grant only the LTS permissions required for the intended use.
The MCP never prints AK/SK. Cloud errors are normalized with an error category, request ID when available, and a retryability hint.
License
MIT. Huawei Cloud SDK packages retain their respective licenses.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。