Sourcegraph Announces General Availability of Agentic Batch Changes, the AI Agent for Large-Scale Code Changes Across Enterprise Codebases
Now generally available, Agentic Batch Changes plans, executes, and tracks code changes across hundreds or thousands of repositories. Pricing is based on changesets merged.
SAN FRANCISCO, September 14, 2026--Sourcegraph, the code intelligence platform that helps enterprise engineering teams understand, oversee, and evolve their codebases, announced the general availability of Agentic Batch Changes for Sourcegraph Cloud customers. Agentic Batch Changes is the frontier agent for code change at scale, built on the indexing and change-management infrastructure Sourcegraph has run for the world's largest codebases for years. One prompt plans, executes, and tracks a change across hundreds or thousands of repositories.
"Every engineering leader I talk to has a list of changes that need to happen everywhere. Those are the projects that never get staffed," said Dan Adler, CEO of Sourcegraph. "Batch Changes already merges half a million changesets a year for the largest enterprises in the world. Agentic Batch Changes now plans that work itself and adapts to each repository, and it has already merged nearly a thousand changesets for early adopters during our beta. You describe the change once and it keeps working until the pull requests are ready to merge, with the speed and confidence of changing a single repository."
The problem with keeping an entire codebase up to date
Every engineering team eventually faces the same challenge: a change that needs to happen everywhere. A dependency needs upgrading. An API changes, and every caller needs updating. A security vulnerability drops, and the fix needs to reach every exposed repository before the next audit window.
Sourcegraph has been helping the owners of the largest codebases in the world make these changes for years with Batch Changes, which creates pull requests across all affected repositories and tracks their progress until they are all merged.
The AI era has brought a tidal wave of code, and updating all of that code is more difficult than ever. Enterprise codebases have outgrown what any one engineer can hold in their head, and coding agents are adding code faster than teams can read it. The bigger the codebase-wide change, the less anyone can vouch for it. A single migration turns into hundreds of pull requests, and it's nearly impossible to have the confidence to say that the work is done. So teams defer the work. Nobody volunteers to upgrade a framework across 10,000 repositories and disappear for three months. The backlog grows until something like a vulnerability forces the issue, and then patching one service doesn't help if the fix has to reach all of them.