MCP 服务器为 AI 驱动的命令执行、基于插件的工具以及包括同步脚本执行、安全 Python 评估和知识管理在内的高级功能提供了全面的框架。它可以作为独立服务运行，也可以嵌入其他项目中，以提供一致的 API 以调用工具和管理任务。

项目结构

• mcp_tools/ – 插件框架和内置工具，增强并发管理
• server/ – Starlette 服务器实现，支持 HTTP/SSE 端点和 DataFrame UI
• 插件/ – 可扩展插件（Azure DevOps、Git tools、知识索引、Kusto、CircleCI）
• config/ – 环境管理器和配置辅助工具
• utils/ – 核心工具（异步作业、图形接口、内存管理、向量存储、PyEval）
• 脚本/ – 安装、实用脚本和自动化提交工作流程
• assets/ – 图片和文档资源

安装

该项目使用 uv 进行依赖管理。安装依赖：

或者用 pip 在开发模式下安装：

环境设置

配置由 .env 文件控制。

MCP Server

The MCP Server provides a comprehensive framework for AI-powered command execution, plugin-based tools, and advanced features including synchronous script execution, secure Python evaluation, and knowledge management. It can be run as a standalone service or embedded in other projects to expose a consistent API for invoking tools and managing tasks.

Project Structure

• mcp_tools/ – Plugin framework and built-in tools with enhanced concurrency management
• server/ – Starlette server implementation with HTTP/SSE endpoints and DataFrame UI
• plugins/ – Extensible plugins (Azure DevOps, Git tools, knowledge indexing, Kusto, CircleCI)
• config/ – Environment manager and configuration helpers
• utils/ – Core utilities (async jobs, graph interface, memory management, vector store, PyEval)
• scripts/ – Installation, utility scripts, and automated submission workflows
• assets/ – Images and documentation resources

Installation

The project uses uv for dependency management. Install dependencies with:

uv sync

Or install in development mode using pip:

pip install -e .

Environment Setup

Configuration is controlled by .env files. Create one from the template and edit it with your settings:

cp config/env.template .env

Important variables include repository paths (GIT_ROOT), Azure Repo details (AZREPO_ORG, AZREPO_PROJECT, AZREPO_REPO), and optional PRIVATE_TOOL_ROOT for external tool configuration. The environment manager automatically loads .env files from the repository root, current directory, and your home directory.

Access settings in code via:

from config import env_manager
env_manager.load()
root = env_manager.get_git_root()

See docs/config_overview.md for more information.

Running the Server

After installing dependencies and configuring .env, start the server with:

uv run server/main.py

Connect to the SSE endpoint at http://0.0.0.0:8000/sse or use the additional routes in server/api.py. Background job endpoints are documented in docs/background_jobs_api.md.

Docker

A Dockerfile is included for running the server in a container. Build the image with:

docker build -t mcp-server .

Then start the container exposing port 8000:

docker run -p 8000:8000 mcp-server

See docs/docker.md for more details.

Configuration Files

The server loads prompts and tool definitions from YAML files:

• server/prompts.yaml
• server/tools.yaml

Private overrides can be placed in server/.private/ or in a folder pointed to by PRIVATE_TOOL_ROOT. Files are resolved in this order:

1. PRIVATE_TOOL_ROOT
2. server/.private/
3. Defaults in server/

Tool System

Tools are modular plugins registered through mcp_tools. Built-in utilities include:

• Command Executor – Synchronous and asynchronous command execution with run-to-completion support