The dSPACE AURELION Manager MCP server enables AI assistants (e.g. GitHub Copilot) to configure and control dSPACE AURELION simulations through natural language.
The following software must be installed and running on the same machine as the MCP server:
- dSPACE AURELION Manager 26.1 or newer
Optionally, for a developer-focused setup:
- Python 3.12 or newer
uv, the Python package and project manager used to create the environment, install dependencies, and run the server from this checkout. Installuvfrom the official Astral documentation: https://docs.astral.sh/uv/getting-started/installation/
You can either download the AURELION Manager MCP server from the latest GitHub Release (Windows only) or run it from this repository with a uv-based setup. You can easily verify the installation with a quick prompt, for example: "Which version of the AURELION Manager is currently running?".
To install the latest release (Windows only):
-
Download
aurelion-manager-mcp-v<version>-windows-x64.zipfrom the latest GitHub Release. The archive contains the server executableaurelion-manager-mcp.exe. -
Follow the MCP server installation instructions for your MCP client (see documentation for Claude, Codex, GitHub Copilot, ... ). Here is an exemplary installation for GitHub Copilot as a
.vscode/mcp.jsonentry that points to the extracted executable:{ "servers": { "aurelion_manager": { "type": "stdio", "command": "<PATH_TO_EXTRACTED_RELEASE>\\aurelion-manager-mcp.exe", "env": { "AURELION_MANAGER_API_URL": "http://localhost:8584" } } } }
Alternatively, for a developer-focused uv-based setup:
This setup works on both Windows and Linux, as long as uv is installed and the dSPACE AURELION Manager REST-API is reachable from that machine.
-
Clone this repository.
-
Create the project environment and install dependencies from
pyproject.toml:uv sync # minimal uv sync --extra dev # full developer setup
-
Follow the MCP server installation instructions for your respective MCP client (see documentation for Claude, Codex, GitHub Copilot, ... ). Use
uv runto start the server. For a portable workspace-root.mcp.jsonentry that works across Windows and Linux without hardcoded paths:{ "mcpServers": { "aurelion_manager": { "type": "stdio", "command": "uv", "args": [ "run", "python", "-m", "aurelion_manager_mcp.server" ], "env": { "AURELION_MANAGER_API_URL": "http://localhost:8584" } } } }If your MCP client launches outside the repository root, see AURELIONManagerMCP.cmd for Windows or AURELIONManagerMCP.sh for Linux. Useful development commands:
uv run pytest # run tests uv run ruff format src tests # run formatter uv run ruff check src tests # run linter uv build # build a wheel
The server is configured via environment variables:
| Variable | Default | Description |
|---|---|---|
AURELION_MANAGER_API_URL |
http://localhost:8584 |
Base URL of the running AURELION Manager REST API |
Set the variable before starting the server or in your MCP host configuration (see below).
| Tool | Description |
|---|---|
list_vehicles |
List all vehicles in the open project |
get_vehicle |
Get a vehicle by ID |
create_vehicle |
Create a new vehicle with components and sensor instances |
update_vehicle |
Fully replace a vehicle configuration |
delete_vehicle |
Permanently delete a vehicle |
| Tool | Description |
|---|---|
list_sensors |
List all sensors in the open project |
get_sensor |
Get a sensor by ID |
create_sensor |
Create a new sensor (CAMERA/RADAR/LIDAR/ULTRASONIC/INFRARED) |
update_sensor |
Fully replace a sensor configuration |
delete_sensor |
Permanently delete a sensor |
| Tool | Description |
|---|---|
list_scenarios |
List all scenarios in the open project |
get_scenario |
Get a scenario by ID |
create_scenario |
Create a ModelDesk or OpenScenario scenario |
update_scenario |
Fully replace a scenario configuration |
delete_scenario |
Permanently delete a scenario |
get_scenario_parameters |
Get OpenSCENARIO parameter declarations |
get_scenario_validation_result |
Validate an OpenSCENARIO .xosc file |
| Tool | Description |
|---|---|
list_systems |
List all simulation systems in the open project |
get_system |
Get a system by ID |
create_system |
Create a simulation system (ASM or Replay) |
update_system |
Fully replace a system configuration |
delete_system |
Permanently delete a system |
| Tool | Description |
|---|---|
get_simulation_state |
Get the current simulation state and loaded config |
get_simulation_diagnostics |
Get connectivity status of all simulation components |
check_simulation_preconditions |
Check if system+scenario are ready to synchronize |
synchronize_simulation |
Push configuration to all AURELION instances |
reset_simulation |
Reset the synchronization state |
cancel_synchronization |
Cancel an in-progress synchronization |
get_scenario_state |
Get current scenario playback state and time |
control_scenario_state |
Start (Running), stop (Stopped), or reset (Reset) the scenario |
get_synchronization_settings |
Get settings persisted for subsequent synchronizations (26.3+) |
update_synchronization_settings |
Persist settings for subsequent synchronizations (26.3+) |
get_simulation_hooks |
List synchronization hooks and their enabled state (26.3+) |
set_simulation_hooks |
Enable or disable synchronization hooks (26.3+) |
| Tool | Description |
|---|---|
get_current_project |
Get information about the currently open project |
open_project |
Open an AURELION Manager project by folder path |
get_version |
Get the AURELION Manager version |
get_log_file_path |
Get the path to the toolchain log file |
Resources provide read-only access to project data as MCP resources (URI-addressable):
| URI | Description |
|---|---|
aurelionmgr://vehicles |
All vehicles in the open project |
aurelionmgr://vehicles/{vehicle_id} |
A specific vehicle |
aurelionmgr://sensors |
All sensors in the open project |
aurelionmgr://sensors/{sensor_id} |
A specific sensor |
aurelionmgr://scenarios |
All scenarios in the open project |
aurelionmgr://scenarios/{scenario_id} |
A specific scenario |
aurelionmgr://systems |
All simulation systems |
aurelionmgr://systems/{system_id} |
A specific simulation system |
aurelionmgr://simulation/state |
Current simulation state |
| Prompt | Description |
|---|---|
configure_simulation |
Step-by-step guide to configure and start a simulation |
create_sensor_workflow |
Guide to create a sensor and integrate it into a vehicle and system |
The following example prompts show how you can interact with the AURELION Manager MCP server to perform typical tasks:
Open the AURELION Manager project located in "C:/Projects/MyAURELIONProject".Configure my selected system to use the VEOS simulator running on 127.0.0.1.Synchronize the current configuration.What is the state of the current simulation?
1. open_project(project_folder="C:/Projects/MyAURELIONProject")
2. list_systems() → note system ID
3. list_scenarios() → note scenario ID
4. check_simulation_preconditions(system_id, scenario_id)
5. synchronize_simulation(system_id, scenario_id)
6. control_scenario_state(state="Running")
7. get_scenario_state() → monitor progress
8. control_scenario_state(state="Stopped")
All tools return JSON. On failure, the response contains an error object:
{
"error": {
"code": "NO_OPEN_PROJECT",
"message": "No project is currently open ...",
"type": "permanent"
}
}type |
Meaning | Action |
|---|---|---|
permanent |
4xx — bad input or resource not found | Fix the request; do not retry |
transient |
5xx or network error | Retry with backoff |
Common error codes:
| Code | HTTP | Cause |
|---|---|---|
NO_OPEN_PROJECT |
412 | No project open — call open_project first |
HTTP_404 |
404 | Resource with given ID not found |
HTTP_400 |
400 | Invalid request body |
CONNECTION_ERROR |
— | AURELION Manager not reachable |
TIMEOUT |
— | Request exceeded 30 s timeout |
UNSUPPORTED_BY_MANAGER_VERSION |
— | A tool or field needs a newer AURELION Manager |
UNSUPPORTED_MANAGER_VERSION |
— | The connected AURELION Manager is older than 26.1 |
UNKNOWN_MANAGER_VERSION |
— | The AURELION Manager version could not be parsed |
pytestTests use pytest-httpx to mock all HTTP calls — no running AURELION Manager is needed.
To run a specific test file:
pytest tests/test_vehicles.py -vsrc/
└── aurelion_manager_mcp/
├── __init__.py # Package version
├── client.py # Async HTTP client wrapper with error classification
├── versioning.py # AURELION Manager version detection, tool gating, and request body conversion
├── server.py # FastMCP entry point, resources, and prompts
└── tools/
├── schemas/ # Shared and per-version request models, one module per resource
│ ├── common.py # ApiSchema base and models used by several resources
│ ├── vehicles.py
│ ├── systems.py
│ ├── sensors.py
│ ├── scenarios.py
│ └── simulation.py
├── vehicles.py # Vehicle CRUD tools
├── sensors.py # Sensor CRUD tools
├── scenarios.py # Scenario CRUD + validation tools
├── systems.py # System CRUD tools
├── simulation.py # Simulation control tools
├── projects.py # Project management tools
└── general.py # Version and log file tools
tests/
├── conftest.py
├── openapi/ # OpenAPI specs of the supported AURELION Manager versions
├── test_openapi_contract.py # Request models vs. OpenAPI specs, per version
├── test_versioning.py
├── test_vehicles.py
├── test_sensors.py
├── test_scenarios.py
├── test_systems.py
└── test_simulation.py
-
Define an input model (Pydantic
BaseModel) in the relevant tool file underaurelion_manager_mcp/tools/. EveryFieldmust include adescriptionwith a concrete example. -
Write an async tool function that calls
client.api_request(...)and returnsjson.dumps(result, indent=2). -
Set all four annotations in the
register_toolscall:readOnlyHint—Truefor GET-only operationsdestructiveHint—Falsefor additive/non-destructive operationsidempotentHint—Trueif calling multiple times is safeopenWorldHint—False(the AURELION Manager is a closed local domain)
-
Register the tool inside
register_tools(mcp):mcp.tool(annotations=ToolAnnotations(...))(my_new_tool)
-
Write unit tests in
tests/usinghttpx_mockto mock the HTTP responses.
The underlying REST API is documented via OpenAPI at:
http://<aurelion-manager-host>:8584/api/aurelionmanager/v1.0/swagger
For technical questions and issues related to the dSPACE MCP Servers and related GitHub repositories, please open a GitHub issue. As a valued dSPACE customer, you are always welcome to contact dSPACE Support directly.