Noop Launches AI Agent Team Running Claude Code and Codex on Local Machines

Noop released a tool that runs a small team of Claude Code and Codex agents on users' own computers, breaking a one-sentence goal into parallel tasks and reporting results. It asks for human decisions only at key points via notifications with selectable options, learns from past decisions to act more autonomously, and runs on a primary Linux machine plus a user's own free Cloudflare account.
Event version 1.
Key insights
Local multi-agent orchestration
Noop runs a small team of Claude Code and Codex agents on the user's own computers. A one-sentence goal is broken down and executed in parallel, with results reported back.
Decision-only human interaction
The tool contacts the user only when a decision is required, sending a notification with a short list of options answerable in one tap on a phone.
Adaptive autonomy
Noop learns from prior decisions, asking more frequently early on and taking more independent action as its judgments align with the user's.
Self-hosted deployment
The system runs on the user's machines and their own free Cloudflare account, with an always-on Linux machine as the primary node and other computers added as needed. It is accessible via phone, computer, and browser.