Multi-repo environments

A multi-repo environment gives the agent the full workspace. It can inspect all of the repositories involved, make coordinated changes, and open pull requests in each repository it modified.

When to use one#

Use a multi-repo environment when a single logical change genuinely spans separate frontend, backend, infrastructure, or shared-library repositories, and the pieces have to agree with each other to work.

For a change contained in one repository, a single-repo environment starts faster and keeps the review surface smaller.

Current limitation#

Long-running agents are not yet available for multi-repo environments. Scope multi-repo tasks to work that completes in a single session rather than tasks you expect to run for an extended period.

Reviewing the result#

Multi-repo work produces several pull requests that only make sense together. Note the ordering: a consumer merged before the provider it depends on will break.

Because coordinated changes are exactly where a plausible-looking diff can be subtly wrong, this is a good case for plan mode before the agent starts.

Related#

• Cloud agents overview — the single-repo default
• Setup — defining environments