Poco-Agent offers a cloud-based autonomous agent experience, inspired by Manus and powered by Claude Code. It targets developers and power users seeking a secure, user-friendly, and easily configurable alternative to existing solutions like OpenClaw. The core benefit is an out-of-the-box platform for complex task automation with enhanced safety and a modern interface.

How It Works

The system leverages Claude Code within isolated Docker containers, providing a secure sandbox for task execution, dependency management, and file manipulation without impacting the host. A Next.js frontend and FastAPI backend deliver a visually appealing, mobile-adaptive UI. The architecture supports MCP/skills extensions and custom skills for broad adaptability, alongside asynchronous and scheduled task execution for continuous background operation.

Quick Start & Requirements

Installation is streamlined via a ./scripts/quickstart.sh bash script, which guides users through API key input and service configuration. Key requirements include Docker and Python 3.12+. The agent is accessible at http://localhost:3000. External S3/R2 storage integration is supported via docker-compose.r2.yml.

Highlighted Details

• Secure task execution within isolated Docker containers.
• Modern, high-fidelity UI with mobile responsiveness and live product previews.
• Full Claude Code feature replication (Slash Command, Plan Mode, AskQuestion) and MCP/skills extensibility.
• Asynchronous and scheduled task execution for background automation.
• Integrated browser for web research and GitHub repository connectivity.

Maintenance & Community

Community engagement is primarily through a WeChat group. The README does not detail core contributors, sponsorships, or a public roadmap.

Licensing & Compatibility

Licensed under MIT, allowing permissive use, modification, and distribution, including for commercial applications.

Limitations & Caveats

Functionality is contingent on cloud-based Claude Code access and requires API keys. Reliance on Docker may present an adoption hurdle for less experienced users. Detailed performance benchmarks or direct comparisons to Manus are not provided.