Agentty – A drop-in alternative to claude-code, written in C++26. 11.0 MB binary (github.com)

Agentty has emerged as a lightweight, drop-in alternative to Claude Code, implemented in C++26, featuring a mere 11.0 MB binary. This tool enhances AI pair programming directly within terminal environments by supporting multiple AI models like Claude, GPT, and Groq, while boasting a cold start time of under 1 millisecond. Users can quickly authenticate via OAuth or an API key, and initiate coding sessions without any dependency on Node, Python, or extensive installations, highlighting its efficiency and ease of use. The significance of Agentty lies in its ability to deliver seamless integration with various AI tools while maintaining a focus on security and performance. By operating in a sandbox environment on Linux and macOS, it ensures that file operations are confined to designated workspaces and can even function on air-gapped systems using SSH for secure data relay. Its design incorporates features like persistent Agent Skills that adapt based on previous interactions, dramatically increasing code accuracy from approximately 20% to an impressive 85% with curated skills. This innovation streamlines the coding process, making advanced AI support accessible for developers working across diverse platforms and environments.