About Datarim

Datarim structures project work through Claude Code, Codex, and Cursor agents: define requirements, plan, implement, verify the outcome, and preserve lessons.

The name comes from the Latin root behind both datarium ("a store of data, a place for data") and dator ("a giver, one who gives") — not a passive vault of knowledge, but a store that gives the right data at the right moment. A store that gives.

Use it for development, research, documentation, content, and other work that can be completed in stages.

The catalogue contains 19 agents, 29 commands, and 80 skills. Orchestration coordinates roles, /dr-auto drives staged work, /dr-quick reduces preparation for tiny tasks, and Jev advises on mode and model tier. The full /dr-auto route stops after compliance; archiving is separate.

Datarim 4.3.7 lets you choose reply and document languages separately: language.replies and language.artifacts both default to en. Personal and project preferences change without reinstalling; standalone reporting uses the same settings. Native client rules and security retain authority.

Standalone reporting 0.3.1 adds synchronous SessionStart for supported Codex and Claude Code clients; review a new Codex definition through /hooks. Installation checks do not establish activation or the language of the first message. A generated note inside a reply follows artifact language, while its surrounding explanation follows reply language; explicit bilingual requests and verbatim input retain their exceptions.

Datarim for everyone — a short introduction.

In 4.3.7, Human Outcome Reporting connects a readable explanation to the request and evidence. /dr-explain restates an existing result without execution. Standalone reporting can be installed without the framework.