When AI Agents Break Git: How Anthropic Rebuilt Source Control for the Agent Era

Description and use

Anthropic replaced Git with Scion to handle 30,000 merges per day, as Git is designed for human work patterns.

Problem

Anthropic’s AI coding agents produced 30,000 code merges per day, overwhelming git’s design for human‑scale, offline‑first workflows.

Outcome

Scion enabled Anthropic to absorb 30,000 daily merges without git failures, eliminating downtime and scaling CI/CD and code review to agent workloads.

Workflow

• 1. Measure merge volume and traffic; 2. Identify git limitations for continuous commits; 3. Design Scion with connectivity‑first assumptions; 4. Migrate monorepo to Scion; 5. Deploy complementary tools (Antique, agent‑scale search, IAM).

Evidence

In May 2026, Anthropic was merging about 2,500 code changes per day. By September, that number had exploded to over 30,000 per day — a 12× increase in just four months.

Source

Git was designed for human developers. Humans work in bursts: write code for hours, commit, push, go home. Agents work continuously: write code in seconds, commit immediately, never stop.

Source

Anthropic built Scion, an in‑house version control system built specifically for agent‑scale development. They migrated their entire monorepo from git in five months.

Source