Insight into your SysML v2 models, live
A model server for each workspace that keeps the whole model loaded and current. It checks every file as you type, computes values with their units, decides whether requirements are satisfied, and answers SPARQL queries, for your editor, your terminal, your CI pipeline and your AI agents alike.
Get Sysight See How It Works$ sysight check sysight: starting the server ... ready in 0.1 s 1 files: 0 errors, 0 warnings, 0 notes $ sysight evaluate VehicleModel::car.mass --unit kg 1130 kg $ sysight verify VehicleModel::carMass VehicleModel::carMass: satisfied $ sysight query 'SELECT ?part WHERE { ?part a sysml:Part }' answered over the model and its libraries
One server per workspace answers them all, so the editor, the command line and the agent never disagree about the model.
Even unsaved edits are seen by everyone attached.
The Sysight: SysML v2 Insight extension: diagnostics in the Problems view, navigation into the libraries, values and verdicts on hover and inline, and a query panel with sessions.
The sysight command checks, evaluates, verifies, queries, exports and updates. It starts the server on first use, then answers in milliseconds.
A Model Context Protocol server, so Claude Code and other agents read, query, evaluate, verify and change the same live model, and see the diagnostics each change adds or removes.
sysight mcpHeadless in pipelines, with exit codes for errors and violated requirements. Any LSP editor, Neovim among them, connects through sysight attach.
Everything Sysight does is one endpoint of the model server, with the same name and parameters from every client.
Syntax, names that do not resolve, the specification's validation rules, and asserted constraints and requirements that fail. Every file a change affects is re-checked, including files you have not opened.
Evaluate any attribute or expression in the context of an element. Results come in the unit you ask for, checked against its dimension, and appear inline in the editor beside the expressions that produce them.
Constraints, requirements and satisfy statements decided as satisfied, violated or undetermined, with the reason why. A requirement's verdict is visible where it is written, on hover.
SPARQL 1.1 with entailment: asking for parts finds every kind of part. Evaluation functions bring values and verdicts into query results, and the query panel helps you write queries and charts the answers.
Change the model through an update API applied as edits to the files' text, all or nothing. Each change reports its diff and the diagnostics it added and removed, with a dry run to preview it first.
Export the model as textual notation or as RDF (TriG, N-Quads, Turtle), one named graph per file, and import RDF or .kpar archives back as notation. The round trip parses to the same graphs.
Sysight turns VS Code into a SysML v2 workbench that understands the model, not only its syntax.
package VehicleModel { private import SI::*; part def Vehicle { // evaluates to 1130 kg attribute mass :> ISQ::mass = engine.mass + chassis.mass; part engine { attribute mass :> ISQ::mass = 180 [kg]; } part chassis { attribute mass :> ISQ::mass = 950 [kg]; } } requirement def MassLimit { subject vehicle : Vehicle; require constraint { vehicle.mass <= 1500 [kg] } } part car : Vehicle; requirement carMass : MassLimit; // satisfied satisfy carMass by car; }
Sysight is written in Rust and brings the semantic web foundations behind OML Code to SysML v2.
Your .sysml and .kerml files are read with the parser of Open-MBEE's sysml-toolkit. The standard KerML and SysML libraries are built in.
Each file becomes a named graph in an embedded Oxigraph store, typed by OWL ontologies shaped by the notation, so classes are named after keywords. Names re-resolve incrementally as files change.
One server per workspace serves HTTP, LSP and MCP from the same graph: diagnostics, values, verdicts and queries, plus updates written back as notation.
SysML v2 gives systems engineering a precise textual language and a formal semantics. Sysight puts that precision to work: the model is checked, computed and verified continuously, so problems surface while you write, not at the next review.
Because the model lives in a standard knowledge graph, the same facts that drive your editor drive your reports, your pipeline and your AI assistants. An agent that can query, evaluate and verify the model is an agent whose answers you can check.
Prebuilt for macOS (Apple silicon and Intel), Linux (x86-64 and ARM) and Windows (x86-64).
Tell us about your models and how your teams work with them, and we will help you get Sysight running in your editor, your pipeline and your agents.
Get Sysight Explore OML Code