Archinsight has gone through several stages since its first version. It started as a web-based diagramming tool for describing and visualizing software architecture. Over time, it evolved into a modelling tool, but the language implementation remained limited, with all entity types baked directly into its syntax. Another limitation was the need to set up relatively complicated infrastructure just to run the tool.
Version 3.0 changes all that.
The Insight language is now strongly typed and extensible. Users can define their own entities or extend the existing ones instead of being limited to a fixed set of built-in concepts. This makes the model better suited to real systems, where teams usually have their own abstractions, rules and conventions.
The project has also moved away from being primarily a web application. The model is expected to live close to the code, be stored in Git, reviewed in pull requests, validated in CI and used as part of the normal software development lifecycle. The new CLI and VS Code plugin make Archinsight usable locally and easier to integrate into existing development workflows, without requiring complicated infrastructure to run the tool.
The biggest addition is AI support. Archinsight now ships a skill for Codex, Claude or a generic agent, giving them the context they need to create, modify and analyze architecture models using the Insight language and graph queries. The strongly typed language gives the agent a constrained and deterministic environment to work in, while the skill provides the knowledge required to use it.
Starting with this release, Archinsight is licensed under Apache 2.0.
There is still a lot to build, but version 3.0 establishes the direction: architecture modelling as part of the development workflow, available locally and usable by both humans and AI agents.