PiPiClaw provides a compact, locally-run AI personal assistant designed for users seeking a fast, always-online, and customizable AI agent. It addresses the need for a powerful AI that respects local control, integrates seamlessly with various LLM services, and offers extensive extensibility through a vast ecosystem of "skills." The primary benefit is a highly portable, near-instantaneous AI companion that can automate complex workflows across different operating systems.

How It Works

Built on .NET 10.0 and optimized with Ahead-of-Time (AOT) compilation, PiPiClaw achieves an exceptionally small ~2MB package size and rapid startup. It acts as a local gateway, interfacing with any OpenAI-compatible LLM (e.g., Qwen, OpenAI, Azure, DeepSeek) via user-provided API keys and endpoints. The agent executes commands, manipulates files, analyzes images, performs content searches, and manages scheduled tasks directly on the host machine. Its core advantage lies in this local execution model, combined with native integration for discovering and installing over 10,000 "skills" from the Skill-Hub, all controllable via a terminal or an integrated web console.

Quick Start & Requirements

• Prerequisites: .NET 10.0 SDK or higher, an OpenAI-compatible API Key (e.g., OpenAI, Azure OpenAI, Aliyun DashScope, DeepSeek, Moonshot).
• Installation/Run:
  • Clone the repository: git clone https://github.com/anan1213095357/PiPiClaw.git
  • Navigate to the directory and restore dependencies: cd PiPiClaw && dotnet restore
  • Run the application: dotnet run (initial setup prompts for configuration).
  • For a self-contained executable: dotnet publish -c Release -r <platform> --self-contained true (e.g., win-x64, osx-x64, linux-x64).
• Links: QQ Group Chat: https://qm.qq.com/q/94ufOsyhZ6

Highlighted Details

• Extremely lightweight (~2MB package size), single-file, and capable of starting in seconds.
• Native integration with "Skill-Hub" for seamless searching and installation of 10,000+ extensible skills.
• Supports cross-platform command execution, file I/O, image analysis, content searching, and persistent scheduled tasks.
• Features an integrated, responsive web console (default port 5050) for remote control and model configuration.

Maintenance & Community

The project is primarily developed by "奶茶叔叔" with a provided QQ group for community interaction. No specific details on corporate sponsorship, partnerships, or a public roadmap are present in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license is permissive for commercial use and integration into closed-source projects. The application is cross-platform, supporting Windows, macOS, and Linux.

Limitations & Caveats

Users must have the .NET 10.0 SDK installed. API keys should be secured, as appsettings.json is excluded from Git commits. Command execution inherits the current user's permissions, requiring caution. Network connectivity is necessary for LLM interactions. Large file reads are automatically truncated to manage token limits. On Windows, automatic sudo elevation is not supported; the application must be run with administrator privileges for such operations.