Globulario is a software publisher with a presence on GitHub, offering tools aimed at developers working with AI-assisted coding workflows. Its current catalog consists of a single product, Sensei, which is described as architectural memory for AI coding agents. Sensei addresses a practical challenge in modern development environments: as teams increasingly rely on autonomous or semi-autonomous AI agents to write, modify, and refactor code, those agents often lack awareness of a project's deeper design decisions and constraints. Sensei captures and maintains this context in the form of invariants, forbidden fixes, and contracts. Invariants represent the rules and properties of a codebase that must always hold true, regardless of how the code evolves. Forbidden fixes record solutions or workarounds that have been tried and rejected, preventing agents from reintroducing known-bad changes that humans have already ruled out. Contracts define the agreed-upon boundaries and expectations between components, helping agents respect interfaces and commitments when generating or altering code. Together, these mechanisms serve as a persistent knowledge layer that sits alongside a repository, giving AI coding agents the institutional memory they would otherwise lack. Typical use cases include teams running AI pair programmers or automated refactoring pipelines, organizations maintaining large or legacy codebases where undocumented decisions carry significant risk, and development groups seeking to enforce architectural consistency without relying solely on code review. As an open-source-oriented publisher distributing through GitHub, Globulario positions its offering within the developer tooling and AI engineering category, targeting practitioners who want their AI assistants to operate safely within established architectural boundaries rather than making plausible but damaging changes.
Architectural memory for AI coding agents — invariants, forbidden fixes, and contracts.
Details