Reason machines releases aspen, an open-source git server optimized for high-volume AI agent traffic

Reason Machines has made Aspen publicly available, an open-source Git server specifically engineered to handle the intense commit traffic generated by numerous AI agents. This server independently implements the architectural design previously described by OpenAI for its Pando system. Aspen is self-hostable, utilizing plain Git over HTTPS and SSH, with packs stored in S3-compatible object storage and branch history managed by PostgreSQL. The system is designed to scale writes based on branches rather than entire repositories, addressing a bottleneck noted by Reason Machines where traditional servers queue every push. While OpenAI’s Pando reportedly reduced median push latency by almost fivefold and handles nearly a petabyte daily, Aspen streamlines the process by never unpacking a push and allowing pushes to different branches to operate concurrently. Testing showed that on a single 8-core server setup, Aspen sustained 567 pushes per second for five minutes against a repository mirroring git.git. The software is available for building from source using Rust, with deployment options including Docker Compose for teams and Terraform for multi-cloud infrastructure. Reason Machines noted that one agent built Aspen from its first commit to public release in four days, involving 151 merged PRs.