TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Anthropic says it built an in-house version control system called Scion and moved its monorepo from Git to the new system in five months. The company reported that daily merged changes rose from about 2,500 in May 2026 to more than 30,000 in September, putting pressure on its existing source-control setup.
Anthropic says it built Scion, an in-house version control system, and migrated its monorepo from Git to Scion in five months, responding to a sharp increase in the volume of code changes it merges. The company reported averaging about 2,500 merged changes per day in May 2026 and more than 30,000 per day by September, a workload it said was straining developer infrastructure designed for human-scale coding.
In an engineering report, Anthropic said it created Scion alongside other source-control tools, including a code review interface called Antique, code search and an internal access-control system. The report focuses on Scion and frames it as a redesign for a company where many repository users are agents and automated jobs rather than people working offline.
Anthropic described the scale of its repository operations as a key pressure. The history of its main branch is almost 100 GB, while a compressed working copy is more than 25 GB. The company said it carries out around 25 million code checkouts per day for CI, excluding developer environments, agent sandboxes and research jobs. It estimated that fetching those checkouts naively would amount to more than 600 petabytes of Git traffic per day.
The company said it had tried familiar ways to scale Git, including shallow clones, sparse checkouts, internal mirrors and repository snapshots. At its peak, Anthropic operated several hundred internal Git mirrors, according to the report. It said rising merge volumes made those systems harder to run and made it more difficult to keep snapshots current and preserve history-dependent functions such as diff and blame.
Why Anthropic Changed Its Source Control
Scion reflects a change in how Anthropic says its software is produced: coding capacity has grown faster than supporting infrastructure. When a large share of repository activity comes from online agents and automated jobs, requiring every client to hold a full copy of a large repository can create avoidable data movement and operational work.
Anthropic designed Scion around a server that performs operations such as querying history, creating commits, merging and rebasing. Clients keep a bounded local object cache, manage working files and report filesystem changes. The report says this arrangement reduces how much repository data clients need locally, while exposing source-control queries through APIs that automated tools can use without maintaining a local repository.
The reported experience may be relevant to other organizations adopting AI coding tools, but Anthropic’s workload is unusual, and the article does not establish that other companies face the same scale or need the same architecture. The practical significance is a case study in infrastructure choices as automated code production increases—not evidence that Git is generally inadequate.
As an affiliate, we earn on qualifying purchases.
From Git Mirrors to Server-Based Work
Git is designed as a decentralized version control system: each clone can contain the repository and support operations such as inspecting history or checking out revisions while offline. Anthropic said that model works well for most repositories, but its monorepo’s size and activity made full local copies increasingly costly.
The company’s report says it was already using Git in a largely centralized way, with one authoritative repository and many clients that were online by definition. It therefore designed Scion on different assumptions: clients can reach a server, the server can handle repository operations, and the system should scale horizontally. Anthropic also cited established large-repository systems, including Google’s Piper and Meta’s Sapling, as prior art whose lessons informed its approach.
The migration was part of a wider infrastructure effort. Anthropic said it also built tools for reviewing code, searching the repository and controlling access, and developed ways to manage and land large volumes of reviewed changes. The report does not provide a full technical accounting of those systems or the migration’s operational cost.
“The bottleneck, now, is all of the developer infrastructure built for human scale: source control, builds, and CI.”
— Anthropic, in its engineering report
As an affiliate, we earn on qualifying purchases.
Scion’s Results Beyond Anthropic
The report describes Anthropic’s architecture and workload, but does not provide independent performance measurements comparing Scion with Git or detail specific improvements in checkout speed, reliability, infrastructure costs or developer productivity. It also does not say whether Scion will be released publicly or adopted outside Anthropic.
The reported merge counts are company figures, and the report does not define every operational detail behind the count. Anthropic said it does not expect growth in code production to slow soon, but that is a company expectation rather than a confirmed forecast. It remains unclear how Scion behaves under other organizations’ workloads or what trade-offs its online-client model may impose when connectivity is unavailable.
large repository management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Anthropic Plans to Scale
Anthropic’s report presents Scion as part of an ongoing effort to expand infrastructure capacity as code changes increase. The company said each bottleneck it removes has been followed by more demand for merging code, so its next challenge is likely to involve the wider toolchain as well as source control, including builds and continuous integration.
The report does not announce a public release, a deployment beyond Anthropic, or a specific next milestone for Scion. Further details about measured performance, system operation and the results of the migration would help clarify how broadly its design could apply.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Scion?
Scion is Anthropic’s in-house version control system. The company says it built the system and migrated its monorepo from Git to Scion in five months.
Why did Anthropic move away from Git?
Anthropic said its monorepo’s size and high volume of automated checkouts made its existing Git-based setup harder to operate. It designed Scion for clients that are generally online and can ask a server to perform repository operations.
How many code changes does Anthropic merge?
Anthropic reported an average of about 2,500 merged changes per day in May 2026 and more than 30,000 per day by September 2026.
Will Anthropic release Scion publicly?
The report does not say whether Scion will be released publicly or made available to other organizations.
Source: hn
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
