Integrations

You want something other than a human at a keyboard to work with your models — a script, an editor, a CI job, a web UI, or an AI agent. sysml-rs exposes one service layer through several transports: every interface dispatches into the same command registry, so an operation behaves identically whichever door you come in through. Pick the door, and this page points you at its documentation.

Which interface for which job

Interface	It’s for	Page
CLI (sysml)	Terminal and CI workflows: parse, check, export, simulate, verify	CLI workflows
Language server (sysml-lsp-server)	Diagnostics, completion, and navigation in any LSP-capable editor	The language server · VS Code setup
REST / WebSocket / SSE (sysml-api)	Scripting over HTTP, custom frontends, streaming progress and session events	The service API
MCP (sysml-mcp)	AI agents — every service command as a callable tool over stdio	MCP for AI agents
Simulation App	A browser workbench: browse, run, verify, requirements, analyze	The Simulation App

The machine-readable inventory of what all of these can do is the API server’s GET /commands (its MCP twin is the sysml_command_catalog tool); a rendered snapshot is published as the API & MCP catalogue.

Combinations work too: sysml-api --mcp runs the HTTP server and an MCP handler over the same live model, so a human in the Simulation App and an AI agent can share one session — see MCP for AI agents.

Security defaults, in brief

The networked interface is a local development server, and its defaults are deliberately narrow:

• sysml-api binds 127.0.0.1:8080 — loopback only — and warns at startup if you bind wider (twice, if you do it with no auth token set).
• CORS admits loopback origins only (localhost, 127.0.0.1, [::1]); --permissive-cors restores allow-any and belongs behind a trusted proxy, nowhere else.
• Authentication is off by default. Setting SYSML_API_TOKEN gates write and command routes behind Authorization: Bearer <token>; read routes are never authenticated. There is no rate limiting.
• The MCP server has no network surface — stdio only, with the privileges of whatever launched it.

Anyone who can reach the API port can read and modify the loaded model and run simulations on your machine, so keep it on loopback or behind your own authenticating proxy. The full defaults, with the verification behind each claim, are on the service API page; the threat model and private vulnerability reporting are in SECURITY.md.