OpenAI has introduced Pando, a new Git-compatible server backend that reduces median push latency by nearly 80 percent while serving all development in the company's monorepo. Written in Rust and deployed to production in April 2026, Pando maintains the standard Git client and protocol so that human engineers and Codex agents can continue running normal commands like git push and git fetch. The backend was developed to solve severe scaling bottlenecks caused by Codex agents multiplying parallel work, which drove a massive surge in commits and CI fetches that previously caused repository reliability to occasionally fall below 90 percent.
Pando achieves its high performance by dropping traditional assumptions about repository-wide consistency and unique packfile creation per client, using causal consistency and immutable blob storage instead. The system relies on background workers to group packs into geometrically larger covers using an LSM-tree compaction method, while a delivery service called Express caches packs regionally with hit rates exceeding 99 percent. In production, Pando handles nearly a petabyte of Git data per day using only a handful of CPU cores and successfully supports thousands of concurrent pushers without performance degradation.
This infrastructure breakthrough matters because it demonstrates how engineering teams can scale version control systems to support heavy automated agent workloads without sacrificing developer velocity.