Alook provides an open-source, self-hosted platform to create a collaborative AI workforce from local agents. It enables agents to coordinate via email, share memory, and improve over time, acting as a team managed by the user, thereby enhancing productivity and automating complex workflows.

How It Works

Alook orchestrates local AI agents, assigning them roles and email addresses to foster team-like collaboration. Agents run on the user's machine, maintaining codebase privacy while connecting to external tools like email, dashboards, and calendars. The system leverages email for agent-to-agent and human-to-agent communication, with tasks managed via a Kanban board. Its core advantage lies in its email-native, local-first architecture, enabling continuous learning and traceable decision-making.

Quick Start & Requirements

Installation is initiated via npx @alook/app onboard, which guides through setup, runtime detection, and agent deployment. Access the deployed instance at http://localhost:15210 or claim @alook.ai email addresses via alook.ai. The platform is built with Next.js, Cloudflare Workers, and Bun.

Highlighted Details

• Collaboration Framework: Define agent roles and organizational structures for automatic coordination.
• Email-Native Communication: Each agent possesses a dedicated email inbox for seamless interaction.
• Task Management: Integrated Kanban board for task assignment, progress tracking, and autonomous issue resolution.
• Calendar Integration: Agents manage their own schedules, including recurring tasks and reminders.
• Local-First Execution: Agents operate on the user's machine, ensuring codebases remain private and accessible remotely.
• Self-Learning & Traceability: Agents build context from completed tasks, remember decisions, and all actions are recorded for accountability.
• Agent Agnosticism: Supports various AI agents, with more planned (e.g., Claude Code, Codex, OpenCode; Cursor, Hermes, OpenClaw coming soon).

Maintenance & Community

Community engagement is facilitated through a Discord server. Contribution guidelines are available in CONTRIBUTING.md.

Licensing & Compatibility

The project is licensed under Apache-2.0, a permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

Several agent integrations (Cursor, Hermes, OpenClaw) are marked as "Coming Soon." While agents run locally, the architecture diagram indicates a "Hosted Machine" component (likely Cloudflare Workers) for core communication services like email and WebSockets.