mcp-server-chart
A monorepo hosting multiple MCP servers, including a TypeScript chart server that generates AntV visualization specs, a Python chart sibling, and an AWS ECS MCP server for container management.
README
mcp-server-chart (monorepo)
A monorepo hosting several Model Context Protocol (MCP) servers:
| App | Language | Package manager | Path |
|---|---|---|---|
| ts-server | TypeScript / Node | npm workspaces | apps/ts-server |
| py-server | Python | Poetry | apps/py-server |
| aws-mcp-server | Python | Poetry | apps/aws-mcp-server |
apps/ts-server is the original AntV @antv/mcp-server-chart.
apps/py-server is a Python/FastMCP
chart sibling, managed with Poetry.
apps/aws-mcp-server is an AWS ECS MCP server scaffold (FastMCP + boto3).
Layout
.
├── apps/
│ ├── ts-server/ # TypeScript MCP server (npm workspace)
│ ├── py-server/ # Python chart MCP server (Poetry)
│ └── aws-mcp-server/ # AWS ECS MCP server (Poetry, FastMCP + boto3)
├── package.json # workspace root + cross-app scripts
├── .husky/ # git hooks (run at repo root)
└── .github/workflows/ # CI: ts-server + py-server + aws-server jobs
Prerequisites
- Node.js
>=18(forapps/ts-server) - Python
>=3.10and Poetry (forapps/py-server)
Setup
# TypeScript workspace (installs apps/ts-server deps + git hooks)
npm install
# Python app
npm run py:install # -> poetry -C apps/py-server install
Common tasks
Run from the repo root:
| Command | Runs |
|---|---|
npm run ts:build |
Build the TypeScript server |
npm run ts:test |
Test the TypeScript server (Vitest) |
npm run start |
Launch ts-server via the MCP Inspector |
npm run py:install |
poetry -C apps/py-server install |
npm run py:start |
Run the Python server (poetry run chart-py) |
npm run py:test |
poetry run pytest |
npm run py:lint |
poetry run ruff check . |
npm run aws:install |
poetry -C apps/aws-mcp-server install |
npm run aws:start |
Run the ECS server (poetry run aws-mcp-server) |
npm run aws:test |
poetry -C apps/aws-mcp-server run pytest |
npm run aws:lint |
poetry -C apps/aws-mcp-server run ruff check . |
Each app can also be used directly from its own directory — see the per-app READMEs linked above.
Connecting ts-server to aws-mcp-server
ts-server can discover a running aws-mcp-server over the network (opt-in):
# 1. Start the ECS server on a port (Streamable HTTP at /mcp)
npm run aws:start:http # -> http://127.0.0.1:8123/mcp
# 2. Point ts-server at it; on startup it connects, lists the ECS tools, and logs them
npm run ts:build
AWS_MCP_SERVER_URL=http://127.0.0.1:8123/mcp node apps/ts-server/build/index.js -t streamable
# -> "Connected to aws-mcp-server at http://127.0.0.1:8123/mcp; tools: containerize_app, build_and_push_image_to_ecr"
Configure the target with AWS_MCP_SERVER_URL, or just AWS_MCP_SERVER_PORT
(host defaults to 127.0.0.1, path /mcp). If unset, ts-server skips the
connection entirely; if the AWS server is down, ts-server logs a warning and
keeps serving charts.
License
MIT
The monorepo conversion is complete and verified. Here's the summary.
What changed
Layout (apps/ split, as chosen): mcp-server-chart/ ├── apps/ │ ├── ts-server/ ← existing AntV TypeScript server (moved here, git history preserved) │ └── py-server/ ← new Python MCP server (Poetry + FastMCP) ├── package.json ← npm-workspace root + cross-app scripts ├── README.md ← new monorepo README (old one → apps/ts-server/README.md) ├── .husky/ ← git hooks stay at repo root ├── .gitignore ← added Python ignores └── .github/workflows/ ← build.yml now has ts-server + py-server jobs; release.yml paths fixed
ts-server: all TS files git mv'd in (97 renames, history intact). Removed the workspace-level husky/lint-staged (now at root); prepare → prepack so publishing still builds.
py-server (Poetry, PEP 621 pyproject, poetry check passes):
- chart_py/server.py — FastMCP server with an example generate_line_chart tool that posts an AntV spec to VIS_REQUEST_SERVER; extend via register_tools.
- tests/ (pytest), console entry point chart-py, ruff configured.
Root scripts (run from repo root): npm run ts:build / ts:test / start, and py:install / py:start / py:test / py:lint (delegating to poetry -C apps/py-server)
AWS MCP
Dependencies added (the ones main.py needs): fastmcp (its from fastmcp import FastMCP) and boto3 (ECS/AWS work), plus dev: pytest, pytest-asyncio, ruff, boto3-stubs.
Entry point — main() (already had the if name == "main" guard) wired as a console script: [project.scripts] aws-mcp-server = "awslabs.aws_mcp_server.main:main"
Stubs made it runnable, not just importable: register_module/register_proxy/get_config/secure_tool are safe no-ops/pass-throughs (they run at import time when main.py builds the server), while the actual tool functions (create_infrastructure, build_and_push_image_to_ecr, delete_infrastructure, delete_app) raise NotImplementedError when called — that's your TODO surface.
Monorepo wiring: root npm run aws:install | aws:start | aws:test | aws:lint, a aws-server CI job in build.yml, a lint-staged glob, and README table/layout entries.
check the Start the server, then query it with the Inspector's CLI — this is the exact command I just ran successfully: npm run aws:start:http # in one terminal
in another:
npx @modelcontextprotocol/inspector --cli http://127.0.0.1:8123/mcp --transport http --method tools/list
call a tool:
npx @modelcontextprotocol/inspector --cli http://127.0.0.1:8123/mcp --transport http
--method tools/call --tool-name containerize_app --tool-arg app_path=/path/to/app port=8080
1.run npm install to run nodemodules in the root 2.compile frontend and run a.npm run ts:build b.npm run ts:start or node apps/ts-server/build/index.js -t streamable -p 1122 3.start the backend a.npm run aws:start:http
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。