已弃用2025年11月27日
我个人已经将我的努力转移到了一个更通用的基于OpenAPI规范的MCP上:https://github.com/allen-munsch/yas-mcp 此外,这里实际上有一个由perfect发布的官方测试版:https://pypi.org/project/prefect-mcp/
完善MCP服务器
一种模型上下文协议(MCP)服务器实现 完美,使AI助手能够通过自然语言与Prefect进行交互。
备注:Prefect MCP官方服务器可用 这里。这是一个社区实施。
🚀 快速开始
docker compose up📦 安装
pip安装
pip install mcp-prefect来源
git clone https://github.com/allen-munsch/mcp-prefect
cd mcp-prefect
pip install -e .手动运行
PREFECT_API_URL=http://localhost:4200/api \
PREFECT_API_KEY=your_api_key_here \
MCP_PORT=8000 \
python -m mcp_prefect.main --transport http🛠️ 特性
╭────────────────────────────────────────────────────────────────────────────╮
│ │
│ _ __ ___ _____ __ __ _____________ ____ ____ │
│ _ __ ___ .'____/___ ______/ /_/ |/ / ____/ __ \ |___ \ / __ \ │
│ _ __ ___ / /_ / __ `/ ___/ __/ /|_/ / / / /_/ / ___/ / / / / / │
│ _ __ ___ / __/ / /_/ (__ ) /_/ / / / /___/ ____/ / __/_/ /_/ / │
│ _ __ ___ /_/ \____/____/\__/_/ /_/\____/_/ /_____(*)____/ │
│ │
│ │
│ FastMCP 2.0 │
│ │
│ │
│ 🖥️ Server name: MCP Prefect 3.6.1 │
│ 📦 Transport: STDIO │
│ │
│ 🏎️ FastMCP version: 2.12.3 │
│ 🤝 MCP SDK version: 1.14.1 │
│ │
│ 📚 Docs: https://gofastmcp.com │
│ 🚀 Deploy: https://fastmcp.cloud │
│ │
╰────────────────────────────────────────────────────────────────────────────╯
[11/11/25 02:08:06] INFO Starting MCP server 'MCP Prefect 3.6.1' with transport 'stdio' server.py:1495
✅ Initialized successfully
Server: MCP Prefect 3.6.1 1.14.1
🔄 Listing tools...
🎯 FOUND 64 TOOLS:
================================================================================
📂 ARTIFACT (6 tools)
🔧 create_artifact
🔧 delete_artifact
🔧 get_artifact
🔧 get_artifacts
🔧 get_latest_artifacts
🔧 update_artifact
📂 AUTOMATION (7 tools)
🔧 create_automation
🔧 delete_automation
🔧 get_automation
🔧 get_automations
🔧 pause_automation
🔧 resume_automation
🔧 update_automation
📂 BLOCK (5 tools)
🔧 delete_block_document
🔧 get_block_document
🔧 get_block_documents
🔧 get_block_type
🔧 get_block_types
📂 DEPLOYMENT (8 tools)
🔧 delete_deployment
🔧 get_deployment
🔧 get_deployment_schedule
🔧 get_deployments
🔧 pause_deployment_schedule
🔧 resume_deployment_schedule
🔧 set_deployment_schedule
🔧 update_deployment
📂 FLOW (13 tools)
🔧 cancel_flow_run
🔧 create_flow_run_from_deployment
🔧 delete_flow
🔧 delete_flow_run
🔧 get_flow
🔧 get_flow_run
🔧 get_flow_run_logs
🔧 get_flow_runs
🔧 get_flow_runs_by_flow
🔧 get_flows
🔧 get_task_runs_by_flow_run
🔧 restart_flow_run
🔧 set_flow_run_state
📂 LOG (2 tools)
🔧 create_log
🔧 get_logs
📂 OTHER (1 tools)
🔧 get_health
📂 TASK (4 tools)
🔧 get_task_run
🔧 get_task_run_logs
🔧 get_task_runs
🔧 set_task_run_state
📂 VARIABLE (5 tools)
🔧 create_variable
🔧 delete_variable
🔧 get_variable
🔧 get_variables
🔧 update_variable
📂 WORK (13 tools)
🔧 create_work_queue
🔧 delete_work_queue
🔧 get_current_workspace
🔧 get_work_queue
🔧 get_work_queue_by_name
🔧 get_work_queue_runs
🔧 get_work_queues
🔧 get_workspace
🔧 get_workspace_by_handle
🔧 get_workspaces
🔧 pause_work_queue
🔧 resume_work_queue
🔧 update_work_queue
📊 TOTAL: 64 tools across 10 categories💬 交互示例
AI助手可以帮助您:
流量管理
- “显示我的所有流及其上次运行状态”
- “为‘数据处理’部署创建新的流运行”
- “流运行“abc-123”的当前状态如何?”
部署控制
- “暂停‘每日报告’部署的计划”
- “使用新参数更新'etl管道'部署”
基础设施管理
- “列出所有工作池及其当前状态”
- “为高优先级作业创建新的工作队列”
变量和配置
- “创建一个名为'api_timeout'的变量,值为300”
- “显示名称中包含'config'的所有变量”
监控与调试
- “获取上次失败流运行的日志”
- “显示当前正在运行的所有任务运行”
🤖 平台集成
克劳德桌面版
添加到 claude_desktop_config.json:
{
"mcpServers": {
"prefect": {
"command": "mcp-prefect",
"args": ["--transport", "stdio"]
}
}
}光标MCP
{
"mcpServers": {
"prefect": {
"command": "mcp-prefect",
"args": ["--transport", "stdio"]
}
}
}Gemini CLI
gemini config set mcp-servers.prefect "mcp-prefect --transport stdio"风帆/克劳德代码
{
"mcpServers": {
"prefect": {
"command": "mcp-prefect",
"args": ["--transport", "stdio"],
"env": {
"PREFECT_API_URL": "http://localhost:4200/api",
"PREFECT_API_KEY": "your_api_key_here"
}
}
}
}通用MCP客户端
{
"mcpServers": {
"prefect": {
"command": "mcp-prefect",
"args": ["--transport", "stdio"],
"env": {
"PREFECT_API_URL": "http://localhost:4200/api",
"PREFECT_API_KEY": "your_api_key_here"
}
}
}
}🧪 发展
运行测试
pytest tests/ -v从源头构建
git clone https://github.com/allen-munsch/mcp-prefect
cd mcp-prefect
pip install -e .
python -m mcp_prefect