Use this MCP server to run offline Plant Simulation dispatching experiments and to safely control a live Plant Simulation COM model.
Validate scheduling instances and compare/run/reproduce FIFO, SPT, and EDD policies across seeds with reference-oracle cross-checking and persisted experiment records.
Generate live-verified SimTalk templates and inspect/summarize stored experiment runs.
Open and fingerprint .spp models; list/inspect objects; create/delete devices; connect MaterialFlow objects; read/write allow-listed properties and counters.
Control simulations via EventController status/start/stop/reset/run and configure Source generation.
Create, update, and read SimTalk methods; attach Entrance/Exit/Init controls; declare MU attributes.
Use 30 schema-validated JSON-RPC tools with structured error envelopes, model-root confinement, clone-first SHA-256 protection, read-back verification, and a single-STA COM gateway.
Controls Siemens Tecnomatix Plant Simulation models through COM, enabling live model discovery, object and method management, simulation control, property read/write verification, and offline dispatching experiment validation.
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., "@Plant Simulation Research Bridge Compare FIFO, SPT, and EDD on my model with 3 random seeds."
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.
Control Siemens Tecnomatix Plant Simulation from Python and MCP, run reproducible dispatching experiments offline with independent oracle cross-checking, and drive live Plant Simulation COM models through a single-STA safety gateway.
instance.json ─► psbridge (offline) ──► validate_instance / compare_policies / reproduce_run (FakeDispatchEngine + Oracle)
──► generate_simtalk (SimTalk templates) / demo
┌─► psbridge.mcp (stdio) ──► ComExecutor (single STA thread + op lock)
│
└─► SafeOps ───────────────► Win32ComBackend ──► Tecnomatix.PlantSimulation.RemoteControl
│ ├─► model discovery (.Models.<name>)
│ ├─► clone-first SHA-256 original protection
│ ├─► property / table / method read-back verification
│ └─► EventController reset / start / isRunning pollingPlant Simulation exposes its automation interface (Tecnomatix.PlantSimulation.RemoteControl) as an out-of-process COM server with strict Single-Threaded Apartment (STA) affinity and asymmetric SimTalk conventions:
Creating objects requires class-receiver syntax (<class>.createObject(<frame>, x, y)), and enumeration uses NumNodes / node(i) rather than NumChildren.
Model roots are discovered from inside the loaded .spp file (.Models.<ModelName> via .Models.node(i).Name), not guessed from the filename.
Information-flow classes require the & prefix (.InformationFlow.&Method, &TableFile), multi-line methods load via Method.load, links are created via .MaterialFlow.Connector.connect(a, b), for loops close with next, and EventController.start is asynchronous with a startup transition window and a 1-hour default StartStat reset.
PlantSimulationBridge (psbridge) encodes these verified conventions behind typed validation allow-lists, read-back verification, clone-first model protection, and a dedicated STA COM executor — alongside an offline discrete-event scheduling layer so policies, metrics, and experiment pipelines can be tested cross-platform without a live Plant Simulation installation.
Related MCP server: MarketCanvas-Env MCP Server
30 MCP tools over stdio JSON-RPC 2.0: Run python -m psbridge.mcp to expose 9 offline experiment/SimTalk tools and 21 live model/simulation/method tools with strict JSON Schema validation and non-throwing structured error envelopes.
Single-STA COM gateway: ComExecutor marshals every COM call onto one dedicated STA worker thread (CoInitialize / CoUninitialize paired) with operation-level locking so multi-step read-modify-verify sequences never interleave.
Safety boundaries:
model_root directory confinement on model paths.
protect_original + SHA-256 cloning before mutating research .spp files, with auto_save_on_disconnect=False.
Allow-listed classes and per-class attributes with mandatory read-back verification (VerificationFailed on mismatch).
Unconditional SetNoMessageBox(True) on connect to prevent modal COM stalls, and automatic session cleanup (Quit() for backend-owned instances) on exit.
Verified SimTalk dispatch, job-shop & table-driven generators: psbridge.simtalk.dispatch, psbridge.simtalk.jobshop (supporting both uniform and heterogeneous per-operation durations via @.OpDuration), and psbridge.simtalk.templates (FIFO/SPT/EDD table-driven routing, Delivery Table releases, SingleProc/ParallelProc support, verified bit-for-bit against the reference oracle on Plant Simulation 13).
Offline scheduling experiments & RL environment: FakeDispatchEngine, FIFO/SPT/EDD policies, a completely independent reference oracle (tests/test_crosscheck.py), exact time-weighted WIP and tardiness metrics, SchedulingEnv (reset / step / action_mask), and crash-tolerant JSONL result storage.
Help index & COM surface tooling: tools/pshelp_index.py, tools/api_surface.py, and psbridge.apiquery build searchable JSON inventories from local Plant Simulation HTML help files and inspect live COM readability.
Cross-platform offline mode (experiments, validation, oracle cross-check, SimTalk generation, unit suite):
Python 3.10+ (standard library only for runtime; pytest for tests)
Live Plant Simulation COM mode:
Windows 10/11
Licensed Siemens Tecnomatix Plant Simulation (verified against Plant Simulation 13)
pywin32>=306 matching your Plant Simulation COM architecture
git clone https://github.com/AmaneSuzuha000/PlantSimulationBridge.git
cd PlantSimulationBridge
# Offline experiment & MCP layer
python -m pip install -e .
# Full install with pywin32 (Windows COM) and dev tooling
python -m pip install -e .[windows,dev]To point live smoke/verification scripts at your local template .spp model, set PSBRIDGE_TEMPLATE:
$env:PSBRIDGE_TEMPLATE = "C:\Program Files\Siemens\Tecnomatix Plant Simulation 13\Examples\Demo\DeliveryConcept.spp"python -m psbridge.demo
python -m psbridge.demo --jsonCompare FIFO, SPT, and EDD across multiple random seeds with jitter, verify every schedule against the independent reference oracle, and persist reproducible JSONL records under .psbridge-work/.
Launch the server directly or configure your MCP client as shown in mcp.json.example:
python -m psbridge.mcpSmoke-test the subprocess JSON-RPC transport:
python scripts/mcp_stdio_smoke.py$env:PSBRIDGE_TEMPLATE = "<PathToPlantSimulation>\Examples\Demo\DeliveryConcept.spp"
python scripts/verify_contract.py
python scripts/all_tools_smoke.py
python scripts/real_machine_dispatch_smoke.py
python scripts/real_machine_jobshop_smoke.py
python scripts/real_machine_smoke.py --instance .psbridge-work/demo_instance.json --policy fifo --model-root "<PathToPlantSimulation>\Examples\Demo" --model-name DeliveryConcept.spp --execute --confirm-real-machineEvery tool returns a JSON object with "ok": true | false and a typed "error" payload (code, message, details) on failure.
Tool | Purpose |
| Bridge version, available policies/backends, and verification flags |
| Validate scheduling instance schema, structure, and engine compatibility |
| Run policies across seeds, cross-check against oracle, and summarize metrics |
| Run one policy over seeds and append records to the JSONL result store |
| Repeat a seeded run multiple times and verify bit-identical schedules/metrics |
| Render live-verified model, table, dispatch, control, and result SimTalk templates |
| Diff |
| List stored experiment records from the JSONL result store |
| Aggregate counts, validity statistics, and metric summaries across runs |
SafeOps)Tool | Purpose |
| Report COM session state and currently loaded |
| Load a |
| Enumerate objects in a frame using |
| Inspect one object's name and class path |
| Derive a device from the allow-list without overwriting existing names |
| Delete an object (requires |
| Connect two objects via |
| Read a validated object attribute |
| Validate attribute applicability, write safely (seconds for time types), and read back |
| Read |
| Read |
| Set |
| Stop the running controller and report status |
| Reset the EventController without altering model structure |
| Start and block until stopped, honoring startup grace window ( |
| Configure a |
| Create an information-flow |
| Replace a Method's SimTalk source (via |
| Read back a Method's |
| Attach a Method handle to |
| Declare a custom typed attribute on |
PlantSimulationBridge/
psbridge/
schema.py Instance / schedule data contracts and strict validation
errors.py Typed error hierarchy and stable error codes
config.py Runtime configuration (paths, timeouts, safety flags)
tools.py 30 MCP-ready tool handlers + serialized COM session manager
demo.py Offline FIFO / SPT / EDD demonstration CLI
apiquery.py Queryable API inventory loader and CLI
com/
backend.py SimBackend protocol and value types
executor.py Dedicated single-STA COM worker + OperationScope lock
win32.py Win32 COM backend + verified SimTalk snippet builders
fake.py Deterministic in-memory FakeBackend with fault injection
core/
ops.py SafeOps verification, clone protection, and lifecycle layer
validation.py Identifier, path, class/attribute allow-lists, SimTalk quoting
experiment/
engine.py DispatchEngine protocol and S1-S9 scheduling specification
fake_engine.py Discrete-event simulation engine
metrics.py Makespan, flow time, tardiness, and exact integral avg WIP
validity.py Schedule feasibility and oracle agreement gates
runner.py Reproducible seeded batch runner
store.py Append-only fsynced JSONL result store
scheduling/
policies.py FIFO, SPT, EDD dispatching rules
reference.py Independent reference schedule oracle
env.py RL-compatible reset / step / action_mask environment
simtalk/
dispatch.py Live-verified scheduled source and exit-control routing builders
jobshop.py Live-verified buffered multi-stage job-shop generator
model_spec.py Table-driven model specification types
templates.py Live-verified table-driven SimTalk dispatch & control generators
mcp/
server.py stdio JSON-RPC 2.0 MCP server
scripts/ Contract checks and live smoke harnesses
tools/ Help HTML indexer and live COM attribute prober
docs/ Architecture notes and live verification checklist
tests/ Comprehensive offline suite + opt-in real COM suite# Lint and offline unit/property/cross-check suite (works on Linux & Windows)
python -m ruff check .
python -m pytest -qTo include the real Plant Simulation COM test case on Windows (uses a temporary copy of PSBRIDGE_TEMPLATE):
$env:PSBRIDGE_REAL_COM = "1"
$env:PSBRIDGE_TEMPLATE = "<PathToPlantSimulation>\Examples\Demo\DeliveryConcept.spp"
python -m pytest -qVerified COM & SimTalk scope: All COM lifecycle calls (Dispatch, LoadModel, SaveModel, CloseModel, NewModel, Quit, GetValue, SetValue, ExecuteSimTalk, SetNoMessageBox), dynamic model/controller discovery, .MaterialFlow.Connector.connect, Method.load, TableFile [col, row] cell round-trips, psbridge.simtalk.dispatch, psbridge.simtalk.jobshop (uniform & heterogeneous per-operation durations), and psbridge.simtalk.templates (FIFO/SPT/EDD exact schedule parity + ParallelProc) are verified against Siemens Tecnomatix Plant Simulation 13.
NewModel() addressing: On Plant Simulation 13, a freshly created NewModel() frame is not addressable by .Models COM paths even after SaveModel(). Open and clone an existing template .spp file instead.
Offline dispatching engine scope: FakeDispatchEngine and reference_schedule target deterministic single-capacity machines without flexible multi-candidate routing (EngineUnsupported is raised cleanly if passed capacity > 1 or multi-candidate operations offline, while psbridge.simtalk.templates maps capacity > 1 machines to .MaterialFlow.ParallelProc).
COM session isolation: GetActiveObject does not attach to a user's GUI Plant Simulation session on typical Windows setups; Dispatch starts a dedicated hidden instance owned and cleaned up by Win32ComBackend.
MIT — see LICENSE and THIRD_PARTY_NOTICES.md.
Return bridge version, available policies/backends, and verification status.
Inputs: no named parameters.
Run FIFO/SPT/EDD (or a subset) over seeds and compare standard metrics with reference cross-check.
Inputs: instance (required), seeds, jitter, horizon, +2 more.
Aggregate stored run records (counts by status/policy, metric aggregates, validity failures).
Inputs: work_dir.
Generate model/table/dispatch/result SimTalk templates for Plant Simulation (NOT_VERIFIED).
Inputs: policy (required), instance (required).
List stored run records under a work directory.
Inputs: work_dir.
Create a Method object in a frame and write its SimTalk source.
Inputs: name (required), source (required), work_dir, frame_path, +1 more.
Read a Method's SimTalk source.
Inputs: path (required), work_dir, model_root.
Replace a Method's SimTalk source; the write is read back and compared.
Inputs: path (required), source (required), work_dir, model_root.
Connect a source to a target with a MaterialFlow connector.
Inputs: source (required), target (required), work_dir, frame_path, +1 more.
Create a device from the measured class allow-list and return its path.
Inputs: name (required), class_name (required), x, y, +3 more.
Delete one object. Requires confirm=true; never deletes the frame.
Inputs: target (required), confirm (required), work_dir, frame_path, +1 more.
Read one writable attribute of an object.
Inputs: path (required), attribute (required), work_dir, frame_path, +1 more.
List the objects in a frame (name and real class path per object).
Inputs: work_dir, frame_path, model_root.
Return one object's name and real class path.
Inputs: path (required), work_dir, model_root.
Load a .spp model into the running Plant Simulation and fingerprint the file.
Inputs: path (required), work_dir, model_root.
Read StatNumIn, StatNumOut and StatWorkingTime for named objects in a frame.
Inputs: objects (required), work_dir, frame_path, model_root.
Write one attribute declared writable for that object class, then read it back.
Inputs: path (required), value (required), attribute (required), work_dir, +2 more.
Report whether a model session is connected and which model it tracks.
Inputs: work_dir, model_root.
Declare a custom attribute on the MU class (e.g. OpIndex, integer).
Inputs: kind (required), name (required), work_dir, model_root.
Attach a Method as an object's EntranceCtrl, ExitCtrl or InitCtrl.
Inputs: path (required), method (required), control (required), work_dir, +1 more.
Run the same instance/policy/seed several times and verify identical schedules and metrics.
Inputs: policy (required), instance (required), seed, jitter, +2 more.
Run one dispatch policy over one or more seeds and persist reproducible records.
Inputs: policy (required), instance (required), seeds, jitter, +2 more.
Compare the dispatch engine schedule with the independent reference oracle for one policy.
Inputs: policy (required), instance (required).
Reset the EventController without touching the model.
Inputs: work_dir, model_root.
Start and block until the controller stops, polling isRunning.
Inputs: timeout, end_time, work_dir, model_root, +1 more.
Make a Source emit a fixed number of parts, starting at a given time.
Inputs: path (required), units (required), interval (required), start, +3 more.
Start the EventController. Returns immediately (measured: start is asynchronous).
Inputs: end_time, work_dir, model_root, statistics_start.
Report whether the EventController runs, the simulated time and the end period.
Inputs: work_dir, model_root.
Stop the EventController and report where it stopped.
Inputs: work_dir, model_root.
Validate a scheduling instance and report engine compatibility and summary statistics.
Inputs: instance (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 30 tools
Most tools have clearly distinct resource+action targets, and potentially overlapping pairs (simulation_start vs simulation_run, run_experiment vs compare_policies vs schedule_vs_reference) are differentiated by their descriptions. A few boundaries in the experiment group remain fuzzy, e.g. run_experiment vs reproduce_run vs schedule_vs_reference all execute a policy over seeds.
Tools fall into clear families (model_*, simulation_*, method_*), but the pattern within families flips between noun_verb (model_status, simulation_status, experiment_summary) and verb_noun (model_open, simulation_start, validate_instance, list_runs). The unprefixed tools (bridge_info, generate_simtalk, object_set_control, mu_create_attribute) add further variation, though everything remains readable.
30 tools is heavy relative to the apparent scope, and several (model_get_property/model_object_info, model_status/bridge_info) could plausibly be consolidated. The surface feels like it grew by adding one tool per capability rather than being scoped deliberately.
Scheduling (validate/run/reproduce/compare/list/aggregate) and model manipulation cover strong lifecycle depth, but there are gaps: no way to close or save a model session, no method delete or method listing, and deletion is limited to devices via model_delete_device. Agents can work around these but will hit dead ends when cleaning up or discovering existing methods.
MCP server for progressive tool usage at any scale (see https://klavis.ai)
Trading MCP server for AI agents, with live market data, account reads and controlled execution.
MCP server exposing the Backtest360 engine API as tools for AI agents.
Ovanek MCP: Model Context Protocol server for Ovanek platform tools; writes are dry-run by default.