Archery MCP Server
# Archery MCP Server
[](https://pypi.org/project/archery-mcp-server/)
[](LICENSE)
将 [Archery](https://github.com/hhyo/Archery) SQL 审核查询平台的 API 封装为 [MCP](https://modelcontextprotocol.io/) 工具,让 AI Agent(Kiro/Claude/Cursor 等)能够直接操作 Archery 平台。
## 一键安装(推荐)
作为 [Agent Skill](https://skills.sh/) 安装,自动配置 MCP:
```bash
npx skills add ckall/archery-mcp-server
```
安装后在 MCP 配置中填入你的 Archery 连接信息即可使用。
## 手动安装
### 方式一:uvx(无需预装,推荐)
```json
{
"mcpServers": {
"archery": {
"command": "uvx",
"args": ["archery-mcp-server@latest"],
"env": {
"ARCHERY_BASE_URL": "http://archery.example.com",
"ARCHERY_USERNAME": "admin",
"ARCHERY_PASSWORD": "your_password"
}
}
}
}
```
### 方式二:pip install
```bash
pip install archery-mcp-server
```
然后配置 MCP:
```json
{
"mcpServers": {
"archery": {
"command": "archery-mcp",
"env": {
"ARCHERY_BASE_URL": "http://archery.example.com",
"ARCHERY_USERNAME": "admin",
"ARCHERY_PASSWORD": "your_password"
}
}
}
}
```
## 环境变量
| 变量 | 必填 | 说明 |
|------|------|------|
| `ARCHERY_BASE_URL` | ✅ | Archery 平台地址 |
| `ARCHERY_USERNAME` | ✅ | 登录用户名 |
| `ARCHERY_PASSWORD` | ✅ | 登录密码 |
## 功能覆盖(26 个工具)
| 模块 | 工具数 | 能力 |
|------|--------|------|
| SQL 工单管理 | 6 | 提交/查看/审核/执行/待办/日志 |
| SQL 审核 | 1 | 语法检查(goInception) |
| SQL 查询 | 6 | 执行查询/浏览资源/表结构/历史/收藏 |
| 实例管理 | 6 | CRUD/资源浏览/按表名查实例 |
| 用户管理 | 4 | CRUD |
| 资源组/权限组 | 4 | 列表/创建 |
详细工具说明见 [SKILL.md](./SKILL.md)。
## 项目结构
```
archery-mcp-server/
├── SKILL.md # Agent Skill 定义(安装即用)
├── pyproject.toml # Python 包配置
├── mcp.json # MCP 配置示例
└── src/archery_mcp/
├── __init__.py
├── client.py # Archery API 客户端(JWT认证)
└── server.py # MCP Server(26个工具)
```
## 开发
```bash
# 克隆
git clone https://github.com/ckall/archery-mcp-server.git
cd archery-mcp-server
# 安装开发依赖
pip install -e .
# 本地运行
export ARCHERY_BASE_URL=http://localhost:9123
export ARCHERY_USERNAME=admin
export ARCHERY_PASSWORD=admin
archery-mcp
```
## 兼容性
- Python 3.10+
- Archery 1.8+(需启用 REST API)
- 支持所有 MCP 兼容的 AI IDE:Kiro、Claude Code、Cursor、Windsurf 等
## License
MIT
TDQS
Scored across 27 tools
Tools are generally distinct across domains (users, groups, workflows, SQL query, instances). Minor overlap exists between sqlquery_resources and instance_resource, both retrieving resources inside an instance, but descriptions clarify different contexts (user-access vs admin config).
Most tools follow a snake_case pattern, but styles vary: some are noun_verb (user_create), while others are noun_noun_verb (resource_group_list) or noun_noun (sqlquery_instances). The verb position is inconsistent, e.g., sqlquery_describe_table places the verb in the middle.
27 tools cover multiple domains (users, groups, workflows, SQL, instances). The count is slightly high but justified by the broad scope of a database management platform. Each tool serves a specific operation without excessive redundancy.
Core workflows (SQL ticket lifecycle, CRUD for users/instances) are covered. Notable gaps: resource groups and auth groups lack update and delete tools; user management lacks role/permission assignment. Some domains feel incomplete.