ridu — Review it. Do it.

The review loop for agent work. Your agent pushes a plan, you review it from your phone, the agent keeps building — live, until every gate is closed.

npx @ridu/cli push plan.md      # publish a doc, get a review URL + QR code
npx @ridu/cli wait --json       # block until the human's verdict arrives
npx @ridu/cli ask "Which DB?" -o "Postgres" -o "SQLite"   # live question, blocks
npx @ridu/cli tasks             # live work state: tasks, gates, decisions
npx @ridu/cli delete <id>       # delete a doc — 30-day trash on ridu.ai/me
npx @ridu/cli init              # wire up Claude Code & Codex (MCP + skill)

Structure your Markdown with ```ridu:task and ```ridu:decision blocks — they become interactive checklists, blocking gates, and decision buttons on the review page, with a progress bar that fills while the agent works.

One page per unit of work. Pushing the same document again appends a version; a different document becomes its own page — so an agent's questions and checkmarks never land on another task's review. Commands without a doc id target the page the current session pushed; --doc <id> continues a page on purpose.

Docs: https://ridu.ai/docs/ · Quickstart: https://ridu.ai/docs/quickstart/

© 2026 Felix Wodlei. Proprietary software — see LICENSE.