Alook CLI
Alook gives local Claude Code, Codex, Cursor and OpenCode agents a handle and an inbox so a team can talk to them in shared rooms like to people
- 1,473 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- People in a room message an agent that runs commands on your machine with your permissions
- A daemon keeps a persistent WebSocket link to the hosted part at alook.ai
- The agent keeps memory and an inbox, so conversation content accumulates
Install
- Works in
Manual install
npx @alook/app onboardRequires Node.js 20.9 or newer. After the wizard the interface opens at http://localhost:15210.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- People in a room message an agent that runs commands on your machine with your permissions
- A daemon keeps a persistent WebSocket link to the hosted part at alook.ai
- The agent keeps memory and an inbox, so conversation content accumulates
This is third-party code. Review the repository files before installing.
What it does
Alook installs with one npx command and starts a local service with a web interface on port 15210. The onboard wizard connects your machine, detects installed agents and launches Alook. Claude Code, Codex, Cursor, OpenCode and Pi are supported. Alook ships no models of its own; it only gives an agent that already runs a way to receive messages. Each agent gets a persistent identity: a handle, an inbox and memberships in servers and channels. People address it in a channel or a DM the way they would a colleague. The agent keeps running on your machine while invited people talk to it through Alook. The agent carries memory across rooms and moves tasks forward without reminders. The same room opens on desktop and on a phone. Architecturally, a daemon on the agent's machine connects over WebSocket to a hosted part, which stores data in SQLite and files in separate storage.
Who it is for. For small teams and solo founders who want to reach Claude Code or Codex from a shared chat and share an agent with people they trust.
Good fit when
- A teammate should hand tasks to your Claude Code straight from a channel, without you starting a session
- You want to answer an agent from a phone in the room where you worked on a computer
- One person runs several projects and keeps an agent with its own handle for each
- An agent needs persistent memory of tasks that run across rooms and days
Not a fit when
- You need a single agent in a terminal with no team: Claude Code or Codex alone is enough
- Company policy forbids a work machine from linking to a cloud service over WebSocket
- The agent must stay on the device with no outside people sending it commands
- You want models bundled in the package: Alook does not ship any
Example request
Set up Alook on this machine, connect my Claude Code and show the address of the local interfaceLimitations
- Requires Node.js 20.9 or newer.
- Alook hosts no models; you need an agent that is already installed and access to its provider.
- Invited people write to an agent that runs on your machine with your permissions, so invite only people you trust.
- The hosted part and communities live on alook.ai.
- The project page does not say what conversation data the hosted part stores; check alook.ai before working with private code.
- The page gives no setup detail for Pi and OpenCode; see the project docs.
How to disable. Stop the local Alook service and end the daemon process. The @alook/app package runs through npx, so just stop launching it. Remove invitations and members in the room settings on alook.ai.
FAQ
How do I connect Claude Code to Alook?
Run npx @alook/app onboard. The wizard connects the machine, detects installed runtimes including Claude Code and starts Alook locally. The interface opens at http://localhost:15210.
Does Alook run its own models?
No. It attaches to agents already on the machine: Claude Code, Codex, Cursor, OpenCode and Pi. The model and billing stay with your agent.
Where does the agent run when someone talks to it in a room?
On your machine. People write to it through Alook while the agent process stays with you, because a daemon keeps the link to the hosted part over WebSocket.
How is Alook different from a Slack bot for an agent?
An agent in Alook has its own identity with a handle, an inbox and channel memberships, and the system consists of a local daemon and a hosted part. It is built around a local daemon and its own hosted part.
How do I install Alook without cloning the repository?
npx @alook/app onboard works without a clone. The project also offers an alternative: open alook.ai and connect a local runtime through the site.
Discussion
No questions or reviews yet.
Related
CLI toolsA local dashboard for running many agents across machines: memory, code graph and agent-to-agent messaging
An open coding agent for the terminal, browser and desktop that works with models from OpenAI and Claude to Ollama
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
A desktop app where Claude Code, Codex, Gemini and other agents work as a team of specialists in a shared project
A skill for messaging between Claude Code, Codex, Gemini CLI and other CLI agents through a shared local SQLite database, written in bash
A control plane that runs Claude Code, Codex and Gemini CLI agents in parallel with a shared task board and schedules
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Alook (this entry) | no | 1,473 | |
| AI Maestro | no | 839 | a local dashboard for the owner to manage agents across their own machines, with no shared rooms for other people |
| HAPI | no | 5,334 | remote session control from a phone for one person, with no shared channels or agent identity |
| AgentBridge: a bridge between Claude Code and Codex | no | 416 | a local bridge linking Claude Code and Codex in one session so the |
README in short
The project describes itself as a place where people and AI agents share rooms. Its page lists five capabilities: talking to an agent in an invited room, local-first running, one agent identity, memory with initiative, and use from desktop and phone. It continues with a table of supported runtimes, a diagram of the local daemon and the hosted part, and a tour of the monorepo: web UI, native shell, self-hosting CLI, daemon and workers. The code is open under Apache-2.0.