AgentSys

AgentSys is a modular runtime that orchestrates AI agents through structured pipelines and gated phases. It handles task selection, branch management, code review, artifact cleanup, CI integration, and deployment—automating the work that happens around code generation. The system includes 24 plugins, 49 specialized agents, and 44 skills, with 3,518 tests and support for Claude Code, OpenCode, Codex CLI, Cursor, and Kiro.

Designed for teams using AI coding tools, AgentSys separates concerns: static analysis and AST parsing handle detection, while LLMs focus on synthesis and judgment. The maker reports this approach uses 77% fewer tokens than multi-agent alternatives and achieves better output quality with lower-tier models. Each agent has a narrow scope, and approval gates ensure code passes review and CI before deployment.