Versions:
Sensei is an architectural memory system for AI coding agents, published by Globulario and currently available at version 1.6.0, with two versions listed in the catalog. Its purpose is to give AI coding agents access to the architectural knowledge that would normally exist only in the heads of senior engineers — invariants, failure modes, forbidden fixes, and design intent — so that automated code changes remain consistent with the established rules of a codebase. The software exposes this knowledge as a queryable graph, which an agent consults before making edits, while a CI gate enforces the same rules after changes are proposed. This dual mechanism supports use cases in teams that rely on AI-assisted development but need safeguards against agents violating architectural constraints, reapplying known-bad fixes, or ignoring contracts between components. Sensei is positioned as a local-first solution: the rules are written as YAML files stored directly in the repository and compiled into a local Oxigraph store. Because everything runs locally, the tool requires no SaaS subscription, no user account, and no uploading of source code to external services, which makes it relevant for organizations with privacy or compliance concerns around code handling. Within a software catalog, Sensei fits the developer tools and AI-assisted software engineering category, particularly at the intersection of code quality enforcement, architectural governance, and agent orchestration. By turning architectural decisions into versioned, machine-readable artifacts that live alongside the source they govern, it allows architectural knowledge to be reviewed, diffed, and enforced through the same workflows as the code itself. Teams adopting AI coding agents can use Sensei to codify senior-level judgment — such as which fixes are forbidden and which invariants must hold — and ensure those constraints are respected both at generation time and at integration time, without sending any code or rules outside their own infrastructure.
Tags: