mcp-bw-modelling-api
Enables AI assistants to explore and query SAP BW/4HANA modeling objects and metadata, such as InfoAreas, InfoObjects, ADSOs, CompositeProviders, and Queries.
README
MCP Server - SAP BW Modeling API
An MCP (Model Context Protocol) server that exposes SAP BW4/HANA modeling objects as tools via FastMCP. Allows AI assistants to explore and query your BW system's metadata: InfoAreas, InfoObjects, ADSOs, CompositeProviders, Queries, and more.
Prerequisites
- Python 3.10+
- Access to a SAP BW4/HANA system with the BW Modeling API enabled (
/sap/bw/modeling/) - A SAP user with appropriate authorizations for the Modeling API
Installation
pip install -r requirements.txt
Configuration
The server is configured via environment variables:
| Variable | Required | Description | Default |
|---|---|---|---|
BW_BASE_URL |
Yes | Base URL of the BW system (e.g. https://bw4hana.example.com) |
— |
BW_USERNAME |
Yes | SAP username | — |
BW_PASSWORD |
Yes | SAP password | — |
BW_CLIENT |
No | SAP client number | 100 |
BW_VERIFY_SSL |
No | Verify SSL certificates (true/false) |
false |
Running the Server
Standalone (stdio transport)
python server.py
With FastMCP CLI
fastmcp run server.py:mcp
HTTP transport
fastmcp run server.py:mcp --transport http --port 8000
MCP Client Configuration
Add this to your MCP client config (e.g. .kiro/settings/mcp.json):
{
"mcpServers": {
"bw-modeling": {
"command": "python",
"args": ["path/to/server.py"],
"env": {
"BW_BASE_URL": "https://your-bw-system.example.com",
"BW_USERNAME": "your_user",
"BW_PASSWORD": "your_password",
"BW_CLIENT": "100"
}
}
}
}
Running from GitHub with uvx (no local clone)
Colleagues can run this server straight from the GitHub repo without cloning it,
using uvx (the runner that ships with uv). This
is the Python equivalent of npx for Node packages.
Prerequisite
Each user needs uv installed (it provides the uvx command). Install it once:
- Windows (PowerShell):
powershell -c "irm https://astral.sh/uv/install.ps1 | iex" - macOS/Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
See the uv install guide
for other options. No manual clone or pip install is needed — uvx fetches,
builds, and runs the server on demand.
Client configuration
{
"mcpServers": {
"bw-modeling": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/felipetomazzi-source/mcp-bw-modelling-api.git@v1.0.0",
"mcp-bw-modeling-api"
],
"env": {
"BW_BASE_URL": "https://your-bw-system.example.com",
"BW_USERNAME": "your_user",
"BW_PASSWORD": "your_password",
"BW_CLIENT": "100"
}
}
}
}
Version pinning (recommended for sharing)
The @v1.0.0 suffix pins to a specific release tag. This matters for teams:
- With a pinned tag, each user's
uvxcache stays valid across day-to-day pushes tomain; it only rebuilds when you publish a new tag. - Without a tag (tracking
main), the cache is invalidated on every push, so users hit a slow cold rebuild (see below) after each change you make.
To ship an update: push to main, create a new tag (e.g. v1.1.0), and have
users bump the @v1.0.0 in their config.
First-launch behavior (important)
On the very first launch on a machine (or the first launch of a new version),
uvx clones the repo and installs its dependencies before the server starts.
This cold build can take longer than the MCP client's startup window, so the
server may show as failed on the first attempt and then connect on a retry.
Notes:
-
Kiro's MCP configuration has no startup-timeout setting, so this cannot be extended via config. If a first launch times out, simply restart the server.
-
To avoid the timeout entirely, pre-warm the cache once in a terminal before first use (this fetches + builds without any client timeout):
uvx --from "git+https://github.com/felipetomazzi-source/mcp-bw-modelling-api.git@v1.0.0" mcp-bw-modeling-apiPress Ctrl+C once it prints the "Starting MCP server ... with transport 'stdio'" banner. Subsequent launches from the client are fast.
Available Tools
| Tool | Description |
|---|---|
check_connection |
Test connectivity to the BW system |
search_bw_objects |
Search for any BW object by name/type |
list_infoareas |
List all InfoAreas (organizational folders) |
list_infoobjects |
List InfoObjects (characteristics, key figures) |
list_adsos |
List Advanced DataStore Objects |
list_composite_providers |
List CompositeProviders (HCPR) |
list_queries |
List BEx/BW Queries |
get_object_details |
Get full metadata for any BW object |
get_infoobject_details |
Get InfoObject properties, incl. navigational vs display attributes and compounding |
get_adso_fields |
Get field list of an ADSO |
get_composite_provider_parts |
Get the source (part) providers of a CompositeProvider |
get_infoarea_contents |
List objects in an InfoArea (flat search) |
get_infoarea_tree |
List objects under an InfoArea via the InfoProvider structure (recursive, reliable) |
get_transformation_details |
Get transformation metadata (source, target, rules) |
get_query_structure |
List a query's characteristics, key figures, and measures (restricted + calculated) |
get_query_filters |
Read a query's filters (fixed/default) and restricted key figures |
get_data_flow |
Show lineage: inbound/outbound transformations and DTPs (with source/target, status, and process chains) |
Supported Object Types
| Code | Object Type |
|---|---|
AREA |
InfoArea |
IOBJ |
InfoObject |
ADSO |
DataStore Object (advanced) |
HCPR |
CompositeProvider |
QUERY |
Query |
TRFN |
Transformation |
DTP |
Data Transfer Process |
MPRO |
MultiProvider |
ODSO |
DataStore Object (classic) |
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。