Agent Data Bridge

Agent Data Bridge

An MCP server that provides data bridging from Spring Boot interfaces and a lightweight Python sandbox for script execution. It enables agents to fetch data as Markdown or Parquet files and perform automated data analysis within a controlled environment.

Category
Visit Server

README

Agent Data Bridge

一个面向 Agent 的“数据桥接 + 轻量沙盒执行”服务,提供两种对外形态:

  • FastAPI HTTP API:用于手工/系统直接调用。
  • MCP Server(SSE Transport):把能力以 Tool 的形式暴露给支持 MCP 的客户端。

目前实际实现包含:

  • 通过两步 OAuth2(token -> query)调用 Spring Boot 接口获取数据(见 src/app/services/springboot_client.py)。
  • 将返回结果中的首个 Markdown 表格解析为 DataFrame;小数据直接返回表格,大数据保存为 parquet 到沙盒目录并返回摘要(见 src/app/main.py)。
  • 提供一个简易 Python “沙盒执行”入口:在指定的沙盒目录中运行脚本,支持超时与输出长度截断(见 src/app/services/sandbox.py)。

运行要求

快速开始

  1. 安装依赖
  • uv sync
  1. 启动 HTTP API(FastAPI)
  • uv run -- uvicorn app.main:app --reload --app-dir src

默认监听 http://127.0.0.1:8000

如果希望“一条命令同时启动 REST + MCP”,可使用:

  • uv run -- python -m app.run_all

1.(可选)启动 MCP Server(SSE)

本项目的 MCP Server 默认监听 0.0.0.0:9000(可通过 MCP_HOST/MCP_PORT 修改),并使用 SSE 传输:

  • SSE 端点:GET /sse
  • 消息端点:POST /messages/

启动命令:

  • Windows(PowerShell,推荐):
$env:PYTHONPATH = "src"
uv run -- python -m app.mcp_server
  • 或者切换到 src 目录运行(一次性):
pushd src
uv run -- python -m app.mcp_server
popd
  • macOS/Linux:
PYTHONPATH=src uv run -- python -m app.mcp_server

配置(.env)

复制 .env.example.env 后按需修改。

当前代码实际会用到的配置(与 .env.example 保持一致):

  • REST_HOST / REST_PORT:REST(FastAPI) 监听地址(用于 Docker 启动与一键启动脚本)。
  • MCP_HOST / MCP_PORT:MCP(SSE) 监听地址。
  • APP_ID / APP_SECRET:Spring Boot OAuth2 client credentials(默认 agent/agent)。

说明:

  • SPRING_BOOT_BASE_URL / SPRING_BOOT_API_PATH 目前在代码中未被使用;/api/fetch 与 MCP 的 fetch_data 都会直接使用传入的 host 参数作为目标地址(见 src/app/services/springboot_client.py)。

HTTP API

健康检查

  • GET /health

返回:{"status":"ok"}

拉取数据并返回摘要

  • POST /api/fetch

请求体:

{
  "host": "http://192.168.10.21:3000",
  "userid": "Admin",
  "sql": "select ...",
  "dataset": "demo"
}

行为(与实现一致,见 src/app/main.py):

  • 解析返回结果中的 data.markdown(Markdown 表格)。
  • Rows <= 15:直接返回完整 Markdown 表格。
  • Rows > 15:保存为 parquet 到 SANDBOX_DIR,并返回字段预览 + 前 5 行。

响应:

{ "message": "..." }

示例(curl):

curl -X POST http://127.0.0.1:8000/api/fetch \
  -H "Content-Type: application/json" \
  -d '{"host":"http://192.168.10.21:3000","userid":"Admin","sql":"select 1","dataset":"demo"}'

运行沙盒脚本

  • POST /api/sandbox/run

方式 A:JSON

{ "filename": "anything.py", "code": "print(123)" }

方式 B:multipart/form-data

  • 上传字段名 file(.py 文件)
  • 或者传 code / filename

返回(与实现一致,见 src/app/services/sandbox.py):

{
  "filename": "script_xxx.py",
  "exit_code": 0,
  "stdout": "...",
  "stderr": "..."
}

MCP Tools

MCP Server 目前提供以下 tools(见 src/app/mcp_server.py):

  • fetch_data(host, userid, sql, dataset) -> str
  • sandbox_run(code, filename=None) -> dict
  • sandbox_list_files() -> str

常见用法:先 fetch_data 生成 parquet 文件名,再用 sandbox_run 执行 Python 读取:

import pandas as pd
df = pd.read_parquet("<file_name>")
print(df.head())

目录说明

  • sandbox_storage/:默认沙盒数据目录(可通过 SANDBOX_DIR 覆盖)。
  • sandbox_storage/_scripts/:沙盒执行时写入的临时脚本目录(自动创建)。

安全与限制

  • 沙盒执行不是强隔离:只是把工作目录固定到 SANDBOX_DIR,并加了超时与输出截断;请勿在不可信输入场景直接暴露到公网。
  • Spring Boot 的 client_id/client_secret 已改为从 .env 读取(APP_ID/APP_SECRET,默认 agent/agent)。

Docker

同时启动 REST + MCP(推荐使用 compose,并把沙盒目录挂载到宿主机):

docker compose up --build

端口:

  • REST: http://127.0.0.1:${REST_PORT:-8000}
  • MCP(SSE): http://127.0.0.1:${MCP_PORT:-9000}/sse

数据卷:

  • ./sandbox_storage -> /app/sandbox_storage

常见问题

Windows 下 ModuleNotFoundError: No module named 'app'

使用 PowerShell 运行 MCP Server 时,请先设置:

$env:PYTHONPATH = "src"
uv run -- python -m app.mcp_server

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured