SAPient MCP
A production-ready MCP server that enables LLMs to automate SAP GUI using RoboSAPiens, supporting both stdio and SSE modes.
README
SAPient MCP
Intelligent SAP GUI automation for AI agents
A production-ready Model Context Protocol (MCP) server that enables LLMs (Claude, Copilot, Cursor, etc.) to automate SAP GUI using the RoboSAPiens library.
Compatible MCP Clients
| Client | Mode | Notes |
|---|---|---|
| Claude Desktop | stdio | Full support |
| Claude Code | stdio | Full support |
| Cursor | stdio or SSE | Full support |
| VS Code (GitHub Copilot) | stdio or SSE | Requires Copilot agent mode |
| Windsurf | stdio or SSE | Full support |
| Any MCP client | SSE/HTTP | Via --port flag |
Prerequisites
| Requirement | Details |
|---|---|
| OS | Windows 10/11 only (SAP GUI is Windows-only) |
| Python | 3.10 or newer |
| SAP GUI | SAP GUI for Windows installed |
| SAP Scripting | Must be enabled on server (RZ11: sapgui/user_scripting=TRUE) AND in SAP Logon client settings |
Enable SAP GUI Scripting (one-time setup)
- Open SAP Logon →
Customize Local Layout (Alt+F12)→Options - Go to
Accessibility & Scripting→Scripting - ✅ Enable scripting
- ❌ Disable "Notify when a script attaches to a running SAP GUI session"
- ❌ Disable "Notify when a script opens a connection"
- Ask BASIS admin to run transaction
RZ11→ setsapgui/user_scripting=TRUE
Installation
# Clone the repo
git clone https://github.com/yourorg/sapient-mcp.git
cd sapient-mcp
# Install (using pip or uv)
pip install -e .
# Or with uv (recommended)
uv pip install -e .
Quick Start
Option A — stdio mode (local clients)
SAPient MCP works with any MCP-compatible client. Pick yours below.
Claude Desktop
Edit %APPDATA%\Claude\claude_desktop_config.json:
{
"mcpServers": {
"sapient": {
"command": "python",
"args": ["-m", "sapient_mcp"],
"env": {
"SAPIENT_MCP_SAPLOGON_PATH": "C:\\Program Files (x86)\\SAP\\FrontEnd\\SAPgui\\saplogon.exe",
"SAPIENT_MCP_CAPS": "screenshot,codegen,advanced",
"SAPIENT_MCP_OUTPUT_DIR": "C:\\sapient_output"
}
}
}
}
Claude Code (CLI)
claude mcp add sapient python -m sapient_mcp \
-- --caps screenshot,codegen,advanced
Or edit ~/.claude/mcp.json / .claude/mcp.json in your project root:
{
"mcpServers": {
"sapient": {
"command": "python",
"args": ["-m", "sapient_mcp", "--caps", "screenshot,codegen,advanced"],
"env": {
"SAPIENT_MCP_SAPLOGON_PATH": "C:\\Program Files (x86)\\SAP\\FrontEnd\\SAPgui\\saplogon.exe",
"SAPIENT_MCP_OUTPUT_DIR": "C:\\sapient_output"
}
}
}
}
Cursor
Go to Cursor Settings → MCP → Add new MCP Server → type command, then enter:
python -m sapient_mcp --caps screenshot,codegen,advanced
Or edit ~/.cursor/mcp.json:
{
"mcpServers": {
"sapient": {
"command": "python",
"args": ["-m", "sapient_mcp", "--caps", "screenshot,codegen,advanced"],
"env": {
"SAPIENT_MCP_SAPLOGON_PATH": "C:\\Program Files (x86)\\SAP\\FrontEnd\\SAPgui\\saplogon.exe",
"SAPIENT_MCP_OUTPUT_DIR": "C:\\sapient_output"
}
}
}
}
VS Code (GitHub Copilot)
Install the MCP server via VS Code CLI:
code --add-mcp '{"name":"sapient","command":"python","args":["-m","sapient_mcp","--caps","screenshot,codegen,advanced"]}'
Or add to .vscode/mcp.json in your workspace:
{
"servers": {
"sapient": {
"type": "stdio",
"command": "python",
"args": ["-m", "sapient_mcp", "--caps", "screenshot,codegen,advanced"],
"env": {
"SAPIENT_MCP_SAPLOGON_PATH": "C:\\Program Files (x86)\\SAP\\FrontEnd\\SAPgui\\saplogon.exe",
"SAPIENT_MCP_OUTPUT_DIR": "C:\\sapient_output"
}
}
}
}
Windsurf
Edit ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"sapient": {
"command": "python",
"args": ["-m", "sapient_mcp", "--caps", "screenshot,codegen,advanced"],
"env": {
"SAPIENT_MCP_SAPLOGON_PATH": "C:\\Program Files (x86)\\SAP\\FrontEnd\\SAPgui\\saplogon.exe",
"SAPIENT_MCP_OUTPUT_DIR": "C:\\sapient_output"
}
}
}
}
Note: In stdio mode, SAPient MCP never writes to stdout or stderr — all logs go to the log file only. This is required for clean JSON-RPC communication across all clients.
Option B — SSE/HTTP mode (remote / CI / multi-user)
Best when the SAP machine is separate from where the AI client runs, or when multiple developers want to share one SAPient instance.
Run the server on your Windows SAP machine:
python -m sapient_mcp --port 8765 --caps screenshot,codegen,advanced
Then point any MCP client at the HTTP endpoint:
{
"mcpServers": {
"sapient": {
"url": "http://sap-windows-machine:8765/mcp"
}
}
}
This works identically in Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, and any other MCP-compatible client.
Option C — Config file
python -m sapient_mcp --config C:\sap_config.json
Environment Variables
All settings can be provided as env vars with the ROBOSAP_MCP_ prefix:
| Variable | Default | Description |
|---|---|---|
ROBOSAP_MCP_SAPLOGON_PATH |
C:\...\saplogon.exe |
Path to SAP Logon executable |
ROBOSAP_MCP_SAP_SERVER |
null |
Server description for auto-connect |
ROBOSAP_MCP_SAP_CLIENT |
null |
SAP client for auto-login |
ROBOSAP_MCP_SAP_USER |
null |
SAP user for auto-login |
ROBOSAP_MCP_SAP_PASSWORD |
null |
SAP password (never logged) |
ROBOSAP_MCP_PORT |
null (stdio) |
Port for SSE mode |
ROBOSAP_MCP_CAPS |
"" |
Comma-separated caps: screenshot,codegen,advanced |
ROBOSAP_MCP_OUTPUT_DIR |
./sap_output |
Screenshots and logs directory |
ROBOSAP_MCP_SCREENSHOT_ON_ERROR |
true |
Auto-screenshot on tool failures |
Available Tools
Core (always loaded)
| Tool | Description |
|---|---|
sap_open |
Launch SAP Logon |
sap_connect_to_server |
Connect to SAP server |
sap_connect_to_running |
Attach to running SAP session |
sap_get_session_info |
Read-only: current state & title |
sap_close |
Close SAP |
sap_execute_transaction |
Run a transaction code |
sap_activate_tab |
Click a tab by label |
sap_get_window_title |
Read-only: current window title |
sap_select_menu_item |
Navigate menu bar |
sap_send_key |
Send keyboard key (Enter, F3, etc.) |
sap_fill_text_field |
Fill a field by its label |
sap_clear_text_field |
Clear a field |
sap_set_checkbox |
Check a checkbox |
sap_unset_checkbox |
Uncheck a checkbox |
sap_select_radio_button |
Select a radio button |
sap_push_button |
Click a button by label |
sap_button_exists |
Read-only: check if button exists |
sap_read_text_field |
Read-only: read field value |
sap_read_text |
Read-only: read any text element |
sap_read_status_bar |
Read-only: SAP status bar message |
sap_count_table_rows |
Read-only: table row count |
sap_select_table_row |
Select a table row |
sap_read_table_cell |
Read-only: read a cell value |
sap_fill_cell |
Fill a table cell |
sap_double_click_cell |
Double-click a table cell |
sap_scroll_table |
Scroll table up/down |
--caps screenshot
| Tool | Description |
|---|---|
sap_take_screenshot |
Capture SAP window screenshot |
--caps codegen
| Tool | Description |
|---|---|
sap_get_generated_script |
Get accumulated Robot Framework script |
sap_clear_script |
Clear script buffer |
--caps advanced
| Tool | Description |
|---|---|
sap_get_snapshot |
Structured JSON snapshot of current window |
Example LLM Conversation
You: Create a purchase order for vendor 100001, company code 1000, with 50 units of material ABC-001
SAPient will guide the LLM to:
- Call
sap_execute_transaction("/nME21N") - Call
sap_get_window_title()→ "Create Purchase Order" - Call
sap_fill_text_field("Vendor", "100001") - Call
sap_fill_text_field("Company Code", "1000") - Call
sap_send_key("Enter")to accept header - Call
sap_activate_tab("Item Overview") - Call
sap_fill_cell("1", "Material", "ABC-001") - Call
sap_fill_cell("1", "Quantity", "50") - Call
sap_push_button("Save") - Call
sap_read_status_bar()→ "Purchase order 4500001234 created"
Project Structure
sapient-mcp/
├── pyproject.toml
├── README.md
├── sap_config.json # Example config file
├── claude_desktop_config.json # Example Claude Desktop config
└── src/
└── sapient_mcp/
├── __init__.py
├── __main__.py # Entry point + CLI arg parsing
├── config.py # Pydantic settings (env/file/CLI)
├── session.py # SAPSessionManager singleton
└── server.py # FastMCP server + all 27 tool definitions
Logs
Logs are written to {output_dir}/sapient_mcp.log.
In stdio mode, logs never go to stdout/stderr (would break JSON-RPC). In SSE mode, logs also appear on stderr.
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.