cowork-to-code-bridge
A bridge that lets Claude Cowork or Claude in the browser hand a task to a real Claude Code agent running on the user's own Mac, Linux, or WSL2 machine, streaming the result back to the chat.
Claude Cowork plans and edits well but runs in a sealed cloud sandbox with no access to a user's actual machine. cowork-to-code-bridge connects the two: Cowork hands a task to Claude Code running locally, which has the real shell, repositories, and tools, and a local agent does the work while the result streams back to the chat, so a request like "run the test suite and fix what's failing" actually executes on the developer's machine.
As a side effect, the same bridge lets Cowork run pre-approved shell scripts directly, without going through the agent, for fixed actions such as builds or disk checks. Tasks are idempotent: a retry after a dropped connection returns the cached result instead of re-running the agent or script.
The bridge opens no network ports, never uses sudo, runs only scripts the user approved, is gated by a secret token, and uninstalls completely with one command; installation is a one-line script for macOS, Linux, and WSL2.