onboard

To generate Claude Code documentation for an existing messy codebase, you can use an onboarding workflow that inspects your repository structure and creates scoped specialist docs like CLAUDE.md and AGENTS.md. This process rehabilitates your project into AI-OS conventions by reflecting the actual code rather than applying generic templates.

Creating a session handoff document for AI agents involves generating a SESSION-HANDOFF.md file alongside durable project constitutions like AGENTS.md. This establishes AI-OS memory scaffolding within directories such as docs/ai-memory and .ai, ensuring continuous context across different development sessions.

You can update AI project documentation without overwriting manual edits by using an idempotent update mechanism. This approach reads existing files first and performs merge-safe appending or merging rather than clobbering user-edited content during repository rehabilitation.

Classifying the project shape of a brownfield repository requires inspecting the existing stack, structure, and presence of auth/DB/API patterns. This discovery process identifies the dominant shape, such as marketing, CRM, e-commerce, SaaS, API, library, monorepo, or mobile app.

Repo onboarding does work for backend APIs and monorepos, alongside marketing sites, CRMs, SaaS, libraries, and mobile apps. The system applies rehabilitation workflows to ensure the generated documentation and routing conventions accurately reflect the real code of these diverse project shapes.