PyAEDT-MCP
Provides tools for driving Ansys Electronics Desktop (AEDT) and PyAEDT: checking for or launching AEDT, connecting to existing gRPC sessions, opening projects, creating designs, running analyses, inspecting models, exporting results, and executing custom PyAEDT Python code — covering solvers such as HFSS, Maxwell 2D/3D, Q2D, Q3D, Icepak, Circuit, TwinBuilder, Mechanical, EMIT, RMXprt, and HFSS 3D Layout.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@PyAEDT-MCP launch AEDT, open my HFSS project, and run the simulation"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
PyAEDT-MCP is an MCP (Model Context Protocol) server for Ansys Electronics Desktop (AEDT). It gives you reliable AEDT tools and a persistent PyAEDT-backed Python session for tasks that do not fit a dedicated tool.
What PyAEDT-MCP does
PyAEDT-MCP supports this runtime loop:
Check whether AEDT is installed or already reachable.
Launch AEDT or connect to an existing gRPC session.
Open projects, create designs, run analyses, inspect the model, and export results.
Use
run_python_codeorrun_python_scriptfor custom PyAEDT work.
Supported AEDT applications include HFSS, Maxwell 2D/3D, Q2D, Q3D, Icepak, Circuit, TwinBuilder, Mechanical, EMIT, RMXprt, and HFSS 3D Layout.
Related MCP server: Ansys MCP Server
Install
Run without cloning
Install uv and
Git first. Git is required because uvx
downloads and runs PyAEDT-MCP directly from this repository without cloning it.
Python 3.13 is recommended; --python 3.13 selects it explicitly and lets uv
download it if needed.
uvx --python 3.13 --from git+https://github.com/ansys/pyaedt-mcp.git ansys-aedt-mcpInstall locally
Use a Python 3.13 or later environment (3.13 recommended).
pip install git+https://github.com/ansys/pyaedt-mcp.gitOr use uv:
uv pip install git+https://github.com/ansys/pyaedt-mcp.gitWindows installer
Windows releases include PyAEDT-MCP-Installer-windows.exe:
Installs the desktop manager into
C:\Program Files\ANSYS Inc\PyAEDT MCP.Adds a Start Menu entry and registers an uninstaller in Windows Installed Apps.
On first use, creates
%APPDATA%\.pyaedt_mcp\.venvand uses bundled CPython 3.13 anduvto installansys-aedt-mcpfrom PyPI.Supports offline installation from the Settings tab. Each install or update prompts you to select a Python 3.13 wheelhouse ZIP and installs only from that archive, without querying package indexes.
Saves the offline installation selection in
%APPDATA%\.pyaedt_mcp\settings.yaml.
Opening the executable provides setup, HTTP server controls, AEDT options, and stdio or HTTP profiles for Copilot/VS Code, Claude, Cursor, Codex, and OpenCode.
Install for development
git clone https://github.com/ansys/pyaedt-mcp.git
cd pyaedt-mcp
pip install -e .
pre-commit installOr use uv:
git clone https://github.com/ansys/pyaedt-mcp.git
cd pyaedt-mcp
uv pip install -e .
pre-commit installRequirements
Python 3.13 or later (3.13 recommended)
Git to install or run PyAEDT-MCP directly from this repository with
uvAEDT 2022 R2 or later for gRPC workflows
A local AEDT installation or a reachable remote AEDT gRPC endpoint
Quick start
Start AEDT in gRPC mode when connecting to an existing session:
"C:\Program Files\ANSYS Inc\v261\AnsysEM\ansysedt.exe" -grpcsrv 50051Or, just launch AEDT normally. It automatically starts a gRPC server.
PyAEDT-MCP can also launch AEDT for you if it is not already running.
Choose an MCP transport:
stdio is the simplest option for one local client. The client starts PyAEDT-MCP automatically through
uvxor from a cloned repository's virtual environment.HTTP connects the client to a separately running server. Use it for local development, Docker, or a remote server.
Start an HTTP server through
uvxwithout cloning:uvx --python 3.13 --index-strategy unsafe-best-match \ --from git+https://github.com/ansys/pyaedt-mcp.git ansys-aedt-mcp \ --transport http --http-host 127.0.0.1 --http-port 8080Point an MCP client at the server:
Use the IDE and client configuration guide for short HTTP and stdio configurations for VS Code/Copilot, Claude, OpenCode, and Codex, plus local virtual-environment instructions.
Tool reference
For information on the available tools, see Tools and capabilities in the PyAEDT-MCP documentation.
How the repository is organized
The core package is in src\ansys\aedt\mcp:
File | Role |
| Module entry point that forwards to the CLI launcher |
| CLI parsing, app setup, context creation, startup cleanup, and transport selection |
| Runtime tool implementations for AEDT lifecycle, project, scripting, and export workflows |
| Small utilities for probing endpoints, normalizing versions, and extracting model data |
| System prompt content shown to MCP clients |
| Optional context helper tools enabled by |
|
|
| Startup code loaded into the persistent Python session |
Other top-level folders include:
doc\source: Sphinx documentationtests: Unit and integration coveragedocker: Container assetsexamples: Sample assets used by the project
Development workflow
To contribute, see Contribute in the PyAEDT-MCP documentation.
Add a new tool
Most tools require a live AEDT connection. Tag those tools with REQUIRES_AEDT_TAG in src\ansys\aedt\mcp\tools.py so they can be hidden until the session exists when dynamic discovery is enabled.
For more information, see Add a new tool in the PyAEDT-MCP documentation.
License
Apache 2.0 license. See LICENSE.
Related projects
Repository provenance · 仓库来源
Independent copy of HNNUYUXUAN/pyaedt-mcp, derived from ansys/pyaedt-mcp. Original authorship and license notices are preserved.
Available Tools
21 toolsanalyze_designAnalyze DesignC
Run analysis on an AEDT design.
Inputs: machine, acf_file, num_gpus, num_cores, +9 more. Returns: result
check_aedt_installedCheck Aedt InstalledA
Check if AEDT is installed on the system.
This tool checks for valid AEDT installations on the system and returns information about available versions.
When running inside a Docker container, the tool probes the remote
gRPC endpoint (AEDT_MACHINE:AEDT_PORT) instead of searching
for a local AEDT installation.
Inputs: no named parameters. Returns: result
check_aedt_statusCheck Aedt StatusA
Check the status of AEDT initialization.
This tool is always reachable, even before a connection has been established.
When no AEDT session is active, it returns a short hint describing how to
establish one (launch_aedt or connect_to_aedt). When a session is
active, it returns full status information.
This makes the tool the recommended pre-flight call to decide whether to
launch_aedt (no active session) or connect_to_aedt (existing
session detected).
This tool retrieves comprehensive information from the connected AEDT instance including version, active projects, designs, and connection details.
Inputs: no named parameters. Returns: result
clear_aedtClear AedtB
Clear the AEDT state by closing all projects.
Inputs: close_projects. Returns: result
connect_to_aedtConnect To AedtA
Connect to an existing AEDT instance via gRPC.
This tool establishes a connection to a running AEDT instance using gRPC.
The AEDT instance must be started with a gRPC server enabled:
ansysedt.exe -grpcsrv <port>
Inputs: port, machine, version, design_name, +2 more. Returns: result
create_designCreate DesignC
Create a new design in AEDT.
Inputs: app_type (required), design_name, project_name, solution_type. Returns: result
disconnect_from_aedtDisconnect From AedtA
Disconnect from the AEDT instance.
This tool detaches the MCP session from the AEDT instance and releases associated resources. It can either keep the AEDT window open or close the AEDT desktop entirely.
Inputs: close_desktop, close_projects. Returns: result
display_htmlDisplay HtmlARead-only
Display self-contained HTML inline or in a detached window.
No AEDT connection is required. Inline scripts and styles are supported, but network requests, external assets, forms, popups, and access to the MCP host are blocked. Embed images as data URLs. This does not execute Python.
In "auto" mode the HTML is shown inline when the client supports MCP
Apps. Otherwise it opens in a lightweight detached window (a chromeless
Edge/Chrome app window, or the default browser) on the machine running
the MCP server. The detached window adjusts to the size of the rendered
HTML unless width or height are given.
Inputs: html (required), title, width, height, +1 more.
export_configExport ConfigC
Export the active design configuration as a JSON file.
Inputs: design, output, project, overwrite. Returns: result
export_resultsExport ResultsC
Export simulation results from AEDT.
Inputs: output_path (required), setup_name, export_type. Returns: result
get_model_infoGet Model InfoC
Get information about the current model/design.
Inputs: design_name. Returns: result
get_pyaedt_logsGet Pyaedt LogsA
Return recent PyAEDT file logs and native AEDT Desktop messages.
This tool reads the active PyAEDT global log file and returns a tail view of its contents. When AEDT is connected, it also retrieves native info and error messages for the active project and design. Optionally filter file log lines using a case-insensitive substring.
Inputs: contains, max_chars, tail_lines. Returns: result
launch_aedtLaunch AedtB
Launch a new AEDT instance.
This tool starts a new AEDT instance using PyAEDT's Desktop class,
or it launches directly into a specific AEDT application session when requested.
The launched instance is automatically connected and stored in the context
for subsequent operations.
Inputs: version, application, non_graphical, confirm_new_session. Returns: result
list_designsList DesignsB
List projects and designs for the connected AEDT instance.
Inputs: project_name. Returns: result
list_projectsList ProjectsA
List all open AEDT projects.
Inputs: no named parameters. Returns: result
open_projectOpen ProjectC
Open an AEDT project.
Inputs: project_path (required), design_name. Returns: result
run_python_codeRun Python CodeB
Execute Python code inside AEDT.
This tool runs inline Python code within the AEDT environment. The code
has access to the connected AEDT instance via the desktop variable.
Inputs: code (required). Returns: result
run_python_scriptRun Python ScriptA
Launch a PyAEDT script file with CPython without opening a terminal.
The script runs asynchronously using the MCP server's Python interpreter and inherits the connected AEDT session through the same environment variables used by the PyAEDT script launcher. Scripts must use PyAEDT; native globals such as oDesktop, oProject, and oDesign are not injected. Standard output and errors are written to the returned log file.
Inputs: script_path (required). Returns: result
save_projectSave ProjectC
Save an AEDT project.
Inputs: save_as, project_name. Returns: result
screenshotScreenshotB
Capture a screenshot of the current AEDT design view.
This tool captures the current design preview as an image. It supports model views, field plots, and mesh visualizations, depending on what's currently displayed in AEDT.
Inputs: path, design, project, plot_type, +2 more.
validate_designValidate DesignA
Validate an AEDT design without starting a solve.
Inputs: design_name, project_name. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 21 tools
Most tools target a distinct resource+action, and descriptions clarify boundaries like launch_aedt vs connect_to_aedt, and check_aedt_installed vs check_aedt_status. Overlap remains in a few spots: list_designs also returns projects (overlapping list_projects), get_model_info overlaps check_aedt_status's project/design reporting, and run_python_script vs run_python_code differ only subtly in execution context.
Nearly all tools follow a clean snake_case verb_noun pattern (launch_aedt, list_projects, create_design, export_results, validate_design). The only deviation is the bare noun 'screenshot', which breaks the otherwise predictable pattern.
21 tools is on the higher side but each maps to a real capability across connection lifecycle, project/design management, script execution, and reporting. The set is somewhat broad, with display_html being peripheral to the core AEDT workflow, but nothing feels redundant filler.
Core workflows (launch/connect, project open/save, design create/validate/analyze, results export, script execution) are covered. However there are notable asymmetries: no create_project or close_project, no delete operations for projects/designs, and export_config has no matching import_config, leaving agents with dead ends for common lifecycle steps.
Maintenance
Related MCP Connectors
Governed data discovery, exact queries, decisions, simulations, and runtime utilities over MCP.
Headless CAD engine & MCP server for 2D/3D drafting, solids, DXF/SVG export, and automation.
Start and manage Reason cloud coding agents, projects and sessions from any MCP client.
Connect any MCP client to MetaTrader 4/5 to read prices, manage positions, and place trades.
Related MCP Servers
- AlicenseCqualityDmaintenanceEnables AI agents to control Ansys Electronics Desktop (HFSS, Maxwell, Q3D, etc.) using MCP tools for simulation automation.10063PolyForm Noncommercial 1.0.0
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to interact with Ansys simulation software (Fluent, MAPDL, Mechanical, Geometry) through the Model Context Protocol.71MIT
- FlicenseNot gradedqualityDmaintenanceA clean MCP server and ACT extension for controlling ANSYS Workbench and ANSYS Mechanical from any MCP client, enabling launching Workbench journals, executing Mechanical Python scripts, and interacting with a running Mechanical session via file queue or socket timer bridge.2-
- AlicenseNot gradedqualityBmaintenanceEnables MCP clients like Codex to control Ansys Workbench and ICEM CFD locally, executing scripts, managing projects, and running simulations.1MIT