AgentGIt

✦ 关于 AgentGIt

AgentGit is a session versioning and collaboration tool for AI agent workflows. It helps people save and manage complete agent sessions in a structured history, similar to the way Git manages versions of code. Every conversation turn becomes a traceable version, while each session can be organized as a separate branch. This allows users to return to an earlier point, resume a project with its original context, or explore a different direction without starting over. Core features include saving, inspecting, and resuming agent sessions; creating a new branch from an earlier turn; syncing work across devices; handing an active session to another person or agent; and sharing a complete conversation through a read-only web link. Sessions published to the AgentGit Hub can remain private or be made public as reusable workflows and examples. Published history is append-only, making it easier to understand how a project evolved, why decisions were made, and which approaches did not work. AgentGit currently supports native session resume for Claude Code, Codex, and OpenCode. Cursor sessions can be imported for read-only inspection. Users can work through the command line, or copy AgentGit’s installation prompt into a supported AI agent and let the agent guide them through setup. Common use cases include continuing an AI coding or vibe coding project on another day, moving work between computers, preserving prompts and decisions, handing a project to a teammate, branching from an earlier idea, reviewing how an agent reached a result, and turning successful agent workflows into reusable project knowledge. AgentGit is designed for AI developers, vibe coders, independent makers, product managers, startup teams, and anyone who uses AI agents to build projects but does not want to restart every conversation from scratch. People who are unfamiliar with traditional Git can use a supported agent to assist with installation and basic operations.