Type: url
Original: https://www.phodal.com/blog/autodev-multi-agents-code-review/
Status: partial
Summary
Phodal Huang’s November 2025 post describes AutoDev’s agentic code review: a four-step pipeline of static information collection (Git diff, CodeGraph, linters, issues and tests), LLM analysis producing structured findings, a generated modification plan, and automated fixes. It explains the multi-agent architecture in which a main CodeReviewAgent orchestrates analysis sub-agents and a CodingAgent through a SubAgentManager, a ToolRegistry and a ToolOrchestrator.