pyaedt-cli

PyAEDT CLI skill

You have access to the pyaedt CLI to interact with ANSYS Electronics Desktop (AEDT). This skill teaches you how to use it effectively.

Pre-flight check (run first)

Before any CLI command, verify pyaedt is installed:

pyaedt --json version

If the command fails, stop, and instruct the user to install it first using the all installation target to enable full capabilities.

pip install pyaedt[all]

---

Invoking the CLI

Before running any command, resolve the pyaedt binary once:

1. Try pyaedt directly. If it is on the system PATH, use it as-is.
2. If not found, fall back to the venv-local binary in the current working directory:
   • Windows: .venv\Scripts\pyaedt
   • Linux/macOS: .venv/bin/pyaedt

All examples use pyaedt for brevity. Replace it with the resolved path if needed.

Working directory for generated artifacts

Use temporary directories for all exports and generated files unless the user specifies otherwise.

Rules:

• Create a task-specific subdirectory under the OS temp directory
• Use tempfile.mkdtemp(prefix="pyaedt-skill-") (Python), $env:TEMP (Windows), or ${TMPDIR:-/tmp} (Linux/macOS)
• Do not write to the repository or arbitrary folders unless explicitly requested
• Move artifacts from temp only after successful export if the user wants to keep them

Quick reference:

import tempfile

temp_dir = tempfile.mkdtemp(prefix="pyaedt-skill-")

Pass this path to --path or --output in export commands.

Use --json for non-interactive commands

Use --json for every non-interactive command so output stays structured and machine-parseable.

JSON output returns: {"status": "ok", "data": {...}} or {"status": "error", "error": "message"}

Always parse the status field before using data. Never rely on human-readable output.

Exceptions:

• session attach launches an interactive console, so human mode is acceptable.
• doc commands open browser pages and are not usually consumed as structured automation output.

Current CLI shape

Commands are organized as:

pyaedt [--json] <group> <command> [options]

Available groups in the current CLI:

• session
• project
• script
• export
• test-config
• panels
• doc

Top-level commands:

• version
• aedt-versions

Do not invent other command groups or commands that do not exist in the current CLI. If a required operation is not supported by the CLI, report that to the user instead of trying to work around it.

Port-based model

The current CLI is primarily port-based.

• session start launches a new AEDT instance and returns its gRPC port.
• session list discovers running AEDT instances.
• Most operational commands require --port explicitly.
• session attach can also be used to select an existing AEDT session, including by --project or --design, so not every workflow starts from a known port.
• There is no persisted session connect state to reuse later.

Practical implication:

# Discover or start AEDT
pyaedt --json session list
pyaedt --json session start --version 2026.1 --non-graphical

# Then pass --port to the command you want
pyaedt --json project list --port 50051
pyaedt --json run my_script.py --port 50051

Project and design resolution rules

Design-scoped commands use one shared resolution rule set. This is important and should be applied consistently in guidance and examples.

Commands affected:

• export screenshot
• export config
• session attach when used with --project or --design

Resolution rules:

• If both --project and --design are provided, use them directly.
• If only --project is provided, that project must contain exactly one design.
• If only --design is provided, exactly one project must be open.
• If neither is provided, there must be exactly one open project and exactly one design in it.
• If the selection is ambiguous, the command fails and the missing selector must be provided.

When helping a user, do not describe this behavior as "active project" or "active design" fallback unless the command actually uses that behavior. The intended behavior is uniqueness-based resolution.

Command reference

Top-level commands

Session management

Notes:

• session attach can work interactively if --port is omitted and the user selects an instance.
• --project or --design with session attach only works for gRPC instances with a usable port.

Project commands

Port is required for all project commands to specify which AEDT instance to target.

project create behavior:

• --name is for project creation.
• --project + --design + --type is for design creation.
• Do not mix --name with --design / --type.

Supported --type values:

• Hfss
• Maxwell2d
• Maxwell3d
• Q3d
• Q2d
• Icepak
• Circuit
• TwinBuilder
• Mechanical
• Emit
• Rmxprt
• Hfss3dLayout
• MaxwellCircuit

Script commands

Export commands

Both commands are design-scoped and therefore follow the shared project/design resolution rules above.

Documentation

Common workflows

Workflow 1: Discover or start AEDT

# Check for running instances
pyaedt --json session list

# If none are running, start one
pyaedt --json session start --version 2026.1 --non-graphical

Decision rule:

• If session list returns no processes, start AEDT.
• If it returns one process, reuse that port.
• If it returns multiple processes, ask the user which port to use.

Workflow 2: Inspect projects on a port

pyaedt --json project list --port 50051

Use this first when you need to understand which project and design selectors are required.

Workflow 3: Create a project or a design

# Create a project. Use --name ONLY (do NOT add --design or --type)
pyaedt --json project create --name "PatchAntenna" --port 50051

# Create a design in an existing project. Use --project + --design + --type ONLY (do NOT add --name)
pyaedt --json project create --project "PatchAntenna" --design "HFSSDesign1" --type Hfss --port 50051

Workflow 4: Interactive exploration with session attach

Use session attach as the first choice when the user wants to explore, debug, or iterate on a design interactively. It opens an interactive Python console that is already connected to the target AEDT session. desktop is pre-injected, so PyAEDT commands can be issued one by one without writing a full script.

# Attach to a running instance (port optional; prompts if omitted)
pyaedt session attach --port 50051

# Inside the interactive Python console, the session is already live:
# >>> import ansys.aedt.core as aedt
# >>> hfss = aedt.Hfss(new_desktop=False, port=50051)
# >>> box = hfss.modeler.create_box([0, 0, 0], [10, 10, 10], name="MyBox")
# >>> hfss.modeler["MyBox"].color = (255, 0, 0)

Sending multi-line code to the interactive Python console

When the agent sends code to the interactive Python console that spans more than one line (loops, conditionals, function definitions, etc.), pasting it directly causes an IndentationError or SyntaxError because the console treats each line as a separate input.

Always wrap multi-line blocks in exec(), encoding all newlines as \n inside a single string:

# Fails because the console breaks on the loop body
for i in range(3):
    print(i)

# Correct: send as a single exec() call
exec("for i in range(3):\n    print(i)")

Rules:

• Use exec() for any block that contains indented code (loops, if, with, function or class definitions).
• Single-statement lines (imports, assignments, function calls) do not need exec().
• Escape all inner double-quotes as \" if the outer delimiter is ", or use concatenated string literals as shown above.
• Keep exec() calls as short as possible; if the block grows beyond ~10 lines prefer saving a temp script and using run instead.

When to prefer session attach command over run command:

Prefer session attach first. Graduate to run only once the workflow is known and needs to be repeatable or automated.

Working with multiple designs simultaneously

When the user needs to work on more than one design at the same time, open a separate interactive Python console for each design instead of switching designs inside a single console. This keeps every design live and editable without closing or releasing any of them.

# Terminal 1: console for Design A
pyaedt session attach --port 50051 --project "ProjectA" --design "DesignA"

# Terminal 2: console for Design B (same or different project)
pyaedt session attach --port 50051 --project "ProjectA" --design "DesignB"

# Terminal 3: console for a design in a different project
pyaedt session attach --port 50051 --project "ProjectB" --design "DesignC"

Rules:

• Each console holds its own PyAEDT object bound to a specific design. Closing one console does not close the others.
• All consoles share the same AEDT instance (same --port), so changes made in one are visible in AEDT immediately.
• Do not call desktop.release_desktop() or aedt_app.release_desktop() inside any of these consoles unless you intentionally want to shut down AEDT for all open consoles.
• If the designs live in different AEDT instances, use the matching --port for each console.

Workflow 5: Run automation

pyaedt --json run "<temp-dir>/setup_antenna.py" --port 50051

Workflow 6: Generated script execution

Use this workflow whenever a script has been generated and must be run automatically.

# Step 1: discover instances
pyaedt --json session list

# Step 2: if needed, start one
pyaedt --json session start --version 2026.1 --non-graphical

# Step 3: run the script on the chosen port
pyaedt --json run "<temp-dir>/generated_script.py" --port 50051

Rules:

• Always determine the target port first.
• If multiple instances are running, present the full list and ask which port to use.
• Do not tell the user to run the generated script manually if the workflow requires automatic execution.
• Generated scripts intended for an existing AEDT instance should use new_desktop=False.

Workflow 7: Export design data

# If the instance has exactly one project and one design, selectors can be omitted
pyaedt --json export screenshot --port 50051 --path "<temp-dir>/preview.jpg"
pyaedt --json export config --port 50051 --output "<temp-dir>/config.json" --overwrite

# If ambiguous, specify the missing selector(s)
pyaedt --json export screenshot --port 50051 --project "PatchAntenna" --design "HFSSDesign1" --path "<temp-dir>/preview.jpg"

Error handling

Always parse the status field in JSON output before using data:

{"status": "ok", "data": {...}}
{"status": "error", "error": "message"}

If status is "error", report the error message to the user and stop. Do not attempt to use data.

Guidance rules for this skill

• Resolve the pyaedt binary once at the start (PATH first, then venv fallback) and use pyaedt for all subsequent commands.
• Prefer --json for non-interactive operations.
• Do not reference removed commands or groups.
• For project create: use --name + --port to create a project; use --project + --design + --type + --port to create a design. Never combine --name with --project, --design, or --type.
• When a command is design-scoped, apply the shared project/design resolution rules explicitly.
• When the available project/design selection is ambiguous, ask for the missing selector instead of guessing.
• Use project list --port ... to inspect the AEDT state before choosing selectors.
• Use run for real scripts.
• Use session attach as the first choice for interactive, exploratory, or debugging workflows; fall back to run only when the workflow is defined, repeatable, or automated.
• When sending multi-line code (loops, conditionals, function definitions) to a session attach interactive Python console, always wrap the entire block in a single exec("line1\nline2\n...") call. Never paste raw indented blocks directly. They cause IndentationError or SyntaxError.
• When the user needs to work on multiple designs simultaneously, instruct them to open one session attach console per design (one terminal per design). Never switch designs inside a single console to handle multiple designs at once.
• When multiple instances are found, present the full list with port, version, and PID before asking.
• When a generated script must be executed automatically, determine the target port first and then run run --port ....
• Generated scripts intended to reuse an existing AEDT instance should use new_desktop=False.
• Unless the user explicitly asks for another destination, create and use a task-specific temporary folder for exports, generated scripts, screenshots, configuration files, and other intermediate files.

Performance tips

• Use session list before starting AEDT so you do not launch unnecessary instances.
• Use project list --port ... to inspect available projects and designs before attempting a design-scoped export.
• Use --non-graphical when launching AEDT for headless or agent workflows.

Cleanup

When the workflow requires shutting down AEDT, use:

pyaedt --json session stop --port 50051

If all AEDT instances must be terminated:

pyaedt --json session stop --all