AI Change Inspector

AI Change Inspector is an open-source, independent auditing and human-review platform for AI coding agents. AI agents can generate and modify code at remarkable speed, but developers still need a reliable way to understand what changed, identify potential risks, and maintain control over the final result. AI Change Inspector addresses this problem by connecting to coding agents through the Model Context Protocol (MCP). It captures immutable Git diff snapshots and uses Meta Muse Spark to independently analyze changes and generate evidence-backed risk hypotheses. Unlike AI reviews that simply present model-generated explanations, AI Change Inspector independently validates code references against the original diff, allowing developers to navigate directly to the relevant lines. The platform includes an interactive code explorer, persistent review history, human decisions and comments, read-only feedback retrieval for agents, parent-child review comparisons, large-diff batching, transparent audit coverage, and selective recovery of incomplete analysis batches. In a real demonstration using Cotiza, a quotation calculation application, the system highlighted a potential exception-handling issue. Human feedback led the coding agent to investigate and fix a reproducible Decimal exception. The correction was independently verified with 61 passing tests, including six real HTTP integration tests. Built with FastAPI, React, TypeScript, and SQLite, the application runs locally and is distributed under GPLv3. Our principle is simple: Your agent writes the code. An independent model reviews the changes. You make the decision.