Claude Code does the thinking; Hoody does the typing

Your team is on Claude Max. You want Claude Code's reasoning inside your Hoody containers, under your confinement, not on someone's laptop.

Hoody delegates the turn. The claude CLI drives the reasoning on its own login, and every action it proposes is a Hoody tool executed by Hoody: same files, same shells, same deny-list, same confirmation gates, same Landlock sandbox. When Claude Code spawns subagents, they get no native file or shell tools of their own; they act through the same mapped tools, and their tokens count toward the session's usage.

The reviewer sees one session, one transcript, one set of rules, regardless of which model held the pen. The engineer gets the model they pay for. Security gets the sandbox it approved.