July 2024, Revisited: AWS CodeCommit Closes Its Doors and Platform Risk Gets Real

This post is part of our Three-Year Retrospective series: thirty-six posts, one per month, looking back at what actually mattered in software engineering. This one covers July 2024.

In late July 2024, AWS quietly closed CodeCommit — its managed Git hosting service — to new customers. No deprecation blog post, no migration deadline, no keynote mention. Existing customers could keep using it, but new accounts simply couldn't create repositories anymore. The news surfaced through error messages and forum threads before AWS confirmed it, eventually publishing guidance on migrating to other Git providers. CodeCommit wasn't alone: the same wave swept up Cloud9, CodeStar, and a handful of other services in what the community came to call AWS's "quiet deprecations" summer.

A small service, a large signal

Let's be honest: CodeCommit was never a beloved product. It was Git hosting with IAM integration and not much else — no pull request culture to speak of, a sparse UI, slow feature development for years. Most teams using it had chosen it for compliance reasons (keep everything inside the AWS boundary, authenticate with IAM) or by default, because it was sitting right there in the console next to everything else.

That's exactly why the closure stung. The teams on CodeCommit were disproportionately the ones who had optimized for stability and integration over features — and they were the ones now planning an unscheduled migration. The lesson wasn't "CodeCommit was bad." It was that even an AWS service, from the company whose brand was we never turn things off, could go from quietly maintained to quietly closed.

What the migration actually looked like

The good news was that Git made the data portable by design. git push --mirror to a new remote moved the commits, branches, and tags in minutes. Everything around the repository was the hard part: IAM-based access control had to be re-mapped to GitHub or GitLab permissions, CodePipeline source stages needed rewiring, approval rule templates had no direct equivalent, and every script or developer laptop with a git-remote-codecommit credential helper needed updating.

That asymmetry — data is portable, integration is not — became our standard framing for platform risk. Through late 2024 we watched the consolidation happen: teams overwhelmingly landed on GitHub or GitLab, often discovering that the migration they'd been deferring for years took a week and made their CI/CD story simpler.

The questions worth asking before you adopt anything managed

July 2024 gave engineering leaders a concrete checklist for any managed developer service, and we've taught it ever since:

• What's the exit path? Is the data in a standard format (Git, SQL, OCI images), and how much of your workflow lives in proprietary glue around it?
• Is the vendor invested? Look at the release notes for the past two years. CodeCommit's near-silence was the warning sign in plain sight.
• Is this the vendor's business or a checkbox? GitHub's whole company is Git hosting; for AWS it was one tile in a crowded console.
• What would a 90-day migration cost you? If you can't answer roughly, you're carrying unpriced risk.

None of this argued against managed services — it argued for adopting them with eyes open and exits rehearsed.

Looking back from June 2026

The consolidation thesis played out: GitHub and GitLab absorbed the bulk of CodeCommit's diaspora, and "where the repo lives" became even more consequential once agentic coding tools started operating directly on hosted repositories in 2025. And then came the twist nobody had on their bingo card: AWS brought CodeCommit back to general availability in November 2025 — a genuinely rare reversal that, if anything, reinforced the lesson. A service that can close and reopen on the vendor's timetable is a service you plan exits for, in both directions.

If your team is consolidating Git providers or tightening up its workflow after a migration, our Git for Professional Teams course covers the branching, review, and access-control practices that transfer across any host, and AWS Cloud Foundations includes the service-evaluation framework — lifecycle signals, exit costs, integration depth — that the CodeCommit episode made essential.