sf-demo-builder
Provisions isolated SAP SuccessFactors demo orgs and generates two-surface demo scripts for Joule Chat and Joule Desktop/Claude Code.
README
SF Demo Builder
MCP server that provisions isolated SAP SuccessFactors demo orgs on demand, then generates two-surface demo scripts for Joule Chat and Joule Desktop / Claude Code.
What it does
Phase 1 — Design (design_demo_org): Given a company name, industry, country, and business problem, returns a complete org plan: roles, scenario narrative, live vs story capability manifest, and an Agent Hub card.
Phase 2 — Provision (provision_demo_org): Takes the plan and creates everything possible via SF OData in an isolated company code on SFSALES011375 — employees, org hierarchy, compensation history, talent profiles, succession nominations, goals, spot awards, and IAS logins.
Supporting tools:
list_scenarios— browse all 9 business problem scenariosgenerate_agent_card— Joule Agent Hub-style card for a plangenerate_demo_script— two-surface demo script: Joule Chat beats + Joule Desktop / Claude Code agentic beats, per scenario
Auth — XSUAA principal propagation
In HTTP mode (--http), the server validates SAP Bearer JWTs against accounts.sap.com (RS256, JWKS). The caller's email is extracted from the token and the +alias suffix maps to an SF username:
siddhartha.bhattacharya+se.ceo@sap.com → se.ceo@SFSALES011375
This lets Joule Desktop / Claude Code identify which SF user is running the demo and default personas accordingly.
In stdio mode (default, used by Claude Code's MCP integration), auth is bypassed — the local process boundary is the trust boundary.
Setup
pip install -r requirements.txt
cp .env.example .env
# Edit .env — fill in SF_ADMIN_PASS, IAS_CLIENT_SECRET, EMAIL_PREFIX
Run modes
# stdio mode (Claude Code MCP — default)
python3 server.py
# HTTP mode (remote / Joule Desktop HTTP MCP)
python3 server.py --http --port 8000
# Generate a local dev token (for testing HTTP mode without a real SAP token)
python3 generate_token.py --email your.name+se.ceo@sap.com
Register with Claude Code
~/.claude/settings.json:
{
"mcpServers": {
"sf-demo-builder": {
"command": "python3.11",
"args": ["/path/to/sf-demo-builder/server.py"]
}
}
}
Environment variables
| Variable | Default | Description |
|---|---|---|
XSUAA_JWKS_URI |
https://accounts.sap.com/oauth2/certs |
JWKS endpoint for JWT validation |
XSUAA_ISSUER |
https://accounts.sap.com |
Expected JWT issuer |
SERVER_BASE_URL |
http://localhost:8000 |
Public URL of this server |
PORT |
8000 |
HTTP listen port |
SF_ADMIN_USER |
sfadmin@SFSALES011375 |
SF admin username |
SF_ADMIN_PASS |
(required) | SF admin password |
IAS_CLIENT_ID |
(required) | IAS SCIM system admin client ID |
IAS_CLIENT_SECRET |
(required) | IAS SCIM system admin client secret |
EMAIL_PREFIX |
siddhartha.bhattacharya |
Email local-part before +alias |
Industries and scenarios
Industries: retail, tech, manufacturing, healthcare, financial_services, energy
Business problems: mass_hiring, compensation_planning, talent_retention, skills_learning, performance_goals, workforce_planning, succession_prep, pay_equity_deep_dive, onboarding_readiness
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。