Alook

Alook is an open-source web application developed by Alook AI that solves a collaboration gap: AI coding agents like Claude Code, Codex, Cursor, OpenCode, and Pi usually live in one person's terminal, invisible to the rest of the team. Alook brings those locally running agents into shared web rooms where team members can interact with them directly. Crucially, the agent processes always run on your own machine, using the code libraries and tools you have configured, so nothing migrates to the cloud and your data stays private. Agents in Alook have persistent identity and memory, meaning they maintain context between sessions and can advance tasks automatically instead of forcing you to re-explain everything. The platform supports desktop and mobile use with session continuity across devices, multi-room management with per-room permissions, agent inboxes that capture messages arriving while you are away, and drag-and-drop organization. It is free to use, with setup requiring a small terminal command to start the daemon on the machine being connected.

- Share your local Claude Code, Codex, or Cursor agents with teammates in a room

- Keep agent context across sessions so tasks continue without re-briefing

- Create multiple rooms with different permissions for different projects

- Catch messages sent to agents while you were away via their inboxes

- Developer teams that want members collaborating directly with shared agents

- Independent developers managing their agents and collaborating with occasional partners

- Technology enthusiasts comparing different agents and sharing experiences

- Remote teams that need cross-department agent access without cloud migration

Setup assumes technical comfort: connecting a machine requires pasting and running an npx daemon command in your terminal, which is fine for developers but not for non-technical colleagues. Because agents run on your machine with access to your codebase, think carefully about what each room member can reach, and use room permissions deliberately rather than leaving everything open; a shared agent session can expose context you did not intend to share. The experience is bounded by the agents themselves: Alook coordinates and persists, but output quality depends on Claude Code, Codex, or whichever agent you connect. As an open-source project, features, the daemon command, and stability can change quickly, so track releases. Finally, running agents continuously consumes your model provider's tokens, so monitor usage when multiple teammates share one agent.