AVP

AVP is an open source tool designed specifically for managing concurrent AI agents that write and modify code. Unlike traditional Git workflows, which were built around human collaboration and rely heavily on pull requests and merging, AVP offers a system where coding agents can open, write, and land tasks in parallel without the typical bottlenecks or manual conflict resolution. Each agent’s work is proven against the exact current state of the main branch before becoming part of the shared codebase, minimizing failed integrations and ensuring the main branch remains stable.

Through a system of replay and automated testing, AVP integrates agent contributions sequentially, even if several agents touch the same files or code paths. This eliminates common CI slowdowns and reduces the need for human intervention when multiple agents are working simultaneously. The server maintains a precise, ordered history of the main branch, and agents communicate via a REST API, reporting their changes and receiving updates in real time.

Ideal for teams or individuals running many AI coding agents on large repositories, AVP supports existing Git workflows and can be installed in any Git project. It’s designed for high velocity development environments where parallel agent tasks are standard, helping keep ship cycles rapid and integration smooth.

AVP is ideal for engineers and teams using multiple AI coding agents to automate software development, particularly in fast-paced startups or R&D environments. CTOs, lead developers, and IT managers who manage AI-driven workflows in large codebases will benefit from the tool's ability to coordinate and safely integrate concurrent agent changes without human bottlenecks.

Software teams experimenting with scaling up agent-driven code writing or using frameworks such as Claude Code, Codex, and similar will find AVP useful for maintaining stability on the main branch while maximizing throughput. It's also suited for startups that build and ship rapidly and need to avoid the integration issues inherent in traditional Git flows when many automated agents are involved.