Provides automation and control for Dassault Systèmes' CST Studio Suite 2026.2, enabling inspection of CST Design Environment projects, safe model editing via structured parameters and fixed History templates, metasurface Floquet mode and incidence-angle setup, R/T/A/PCR analysis, job and solver management, and result exports to CSV, HTML, and Touchstone.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
A local Python automation core for CST Studio Suite 2026.2, with a CLI
(cst-rf) and a stdio MCP server (cst-rf-mcp) sharing the same service,
schemas, result envelopes and audit trail.
简体中文 · Architecture · Tool catalog · Acceptance status
Development preview (
0.1.0.dev0), not a fully validated unattended simulation product. Registering a tool does not mean that its CST History commands or solver behavior passed live acceptance.
Inspect: discover/connect to a CST Design Environment, check the expected project identity, inspect the model tree, parameters, boundaries and Floquet modes. Saved results and the locally installed CST help can be read offline.
Model safely: copy a source .cst project to an operation-specific scratch,
then use structured parameters and fixed History templates; source projects
are never live write targets. No arbitrary VBA or shell execution is exposed.
Metasurfaces: set Floquet mode counts and incidence angles; calculate
R/T/A and PCR from explicit saved complex channels and their real frequency
axes. Export CSV, HTML and Touchstone with the recorded reference impedance.
A claimed T=0 is an assumption requiring independent complete-backplane
evidence, not an inference from missing transmission data.
Jobs and safety: require confirm=true for mutation and solver actions,
confine writes to CST_WORK_DIR, use a cross-process project lock and
append-only audit. Ambiguous solver outcomes stay unknown and block later
writes rather than being reported as successful stops.
One periodic unit-cell solve, its S-parameter/R/T/A/PCR analysis, Floquet mode
and incidence-angle readback, and selected exports passed local CST 2026.2
acceptance. Finite-array replication, circular-polarization switching,
successful solver stopping, parameter sweeps and the patch-antenna recipe have
not passed complete live acceptance. The last local stop experiment has an
unresolved unknown job; do not treat its later CST SUCCESS as proof of a
successful stop. See the public acceptance summary
and metasurface status.
Related MCP server: cst-mcp
CLI / MCP stdio -> shared ToolSpec registry -> Service (schemas, audit, lock)
-> core/session (single CST worker thread)
-> fixed History / official CST Python API
-> saved-results and workflow adapterscst.interface is used from its original CST installation location;
cst.results.ProjectFile(..., allow_interactive=False) reads saved data
without starting CST. The CLI exposes a subset of the MCP tool catalog.
See architecture for trust boundaries and lifecycle.
Requires Python 3.12, a licensed CST Studio Suite installation for live tools,
and a compatible CST-provided Python library. The repository does not ship
CST binaries, help files, or results. A conventional installation path is
C:\Program Files\CST Studio Suite 2026; set CST_PATH for another location.
conda env create -f environment.yml
$env:CST_PATH = 'C:\Program Files\CST Studio Suite 2026'
$env:PYTHONPATH = "$env:CST_PATH\AMD64\python_cst_libraries"
$env:CST_WORK_DIR = Join-Path $HOME 'Documents\cst-rf-work'
conda run -n cst-rf-mcp cst-rf inspect status --json
conda run -n cst-rf-mcp cst-rf tools list --jsonThe public environment recipe defaults to manual connection and does not
auto-launch CST. Set CST_WORK_DIR to a private writable folder before any
mutation or solve. An existing conda environment may retain older environment
variables: inspect it with conda env config vars list -n cst-rf-mcp before
running; editing environment.yml does not reset an already activated process.
Automatic launch/open/switch/popup handling is an explicit high-risk opt-in,
not a safe installation default. Unknown dialogs are not clicked blindly.
CST_TOOLSET currently does not filter tools and is not an authorization
boundary; configure MCP client permissions separately. The MCP executable is
cst-rf-mcp; do not expose all mutation and solver tools to untrusted agents.
conda run -n cst-rf-mcp ruff check .
conda run -n cst-rf-mcp ruff format --check .
conda run -n cst-rf-mcp mypy src tests
conda run -n cst-rf-mcp pytest -m "not live and not solver"GitHub Actions runs the same offline checks on Windows. Real CST acceptance requires separate opt-in and is never part of the default test suite. Reference code was not incorporated; see provenance and reference audit.
Apache-2.0; see LICENSE. This is an independent project and is not affiliated with or endorsed by Dassault Systèmes.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Connect your First AI Employee workspace to review authorized business records, including contacts, tasks, calls, bookings, forms and reports. Prepare supported changes as an exact preview for approval before applying them. OAuth sign-in is required; available tools depend on account permissions, enabled products and access. Website: https://firstaiemployee.com/ . Setup: https://developer.firstaiemployee.com/mcp/
Simpro MCP server for ChatGPT, Claude and any AI agent: list jobs, create quote, convert quote to job, update job status, look up customer, check schedule, add schedule entry and search catalogue items in your own Simpro account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Simpro yourself; Cotlet does not store your password and you need no Simpro API key. About the official Simpro API: Simpro has a full REST API with OAuth 2.0 for quotes, jobs, schedules, customers and catalogue items, but no official MCP server (checked 17 Oct 2026). Your administrator creates the API credentials in your build, partner apps need approval from Simpro, and all calls share a limit of 10 per second per build. Setup and FAQ: https://cotlet.dev/integrations/simpro-api-mcp-server
Service Autopilot MCP server for ChatGPT, Claude and any AI agent: create job, reschedule visit, send estimate, list invoices, add client and get route in your own Service Autopilot account. Cotlet works in the account through your Cloud Browser, a real Chrome where you sign in to Service Autopilot yourself; Cotlet does not store your password and you need no Service Autopilot API key. About the official Service Autopilot API: Service Autopilot has no public API, developer portal or self-service API keys (checked 11 Oct 2026). Data moves through a two-way QuickBooks sync, a few partners the vendor picks, a Zapier trigger that only sends completed jobs out, and manual import and export; the QuickBooks sync needs the Pro or Pro Plus plan. Setup and FAQ: https://cotlet.dev/integrations/service-autopilot-api-mcp-server
Hosted email MCP for your own Gmail, Outlook.com, Microsoft 365, iCloud or IMAP inbox: read, search, draft, reply in thread, forward and file mail. It moves or flags up to 500 messages in one call, and a send leaves exactly one copy in Sent. A calendar is a separate connection, and connecting one adds diary and scheduling tools.
