Cursor Adds Multi-Repo Support to Cloud Agent Environments

Cursor has expanded its Cloud Agent Development Environments to support multi-repo workflows, bringing standard engineering tools to autonomous AI coding. Released on May 13, 2026, the update equips cloud-hosted agents with Dockerfile-based configuration, build secrets, and private credential access. This moves the IDE’s automation capabilities past single-file autocomplete and into delegated engineering across complex, interconnected microservices.

The environment upgrades build on the release of Cursor 3 earlier in the year. The company reports that its autonomous agents now generate 30% of internal merged pull requests, supported by underlying foundation architectures like Kimi K2.5.

Multi-Repo and Cross-Service Execution

Previous agent iterations were constrained to single repositories, limiting their ability to handle real-world API changes that require coordinated updates across a frontend and backend. Cloud agents can now mount and modify multiple repositories simultaneously. If an agent alters a database schema in a core service, it can independently update the corresponding types in dependent microservices before submitting the work.

Environment Configuration as Code

Developers define the agent’s workspace using standard Dockerfiles. Cursor has upgraded the underlying image builder, introducing layer caching that makes environment rebuilds 70% faster when hitting the cache after a Dockerfile modification.

Build secrets are now natively supported. Agents can securely access private package registries during the build phase without exposing those credentials in the running environment. If an environment fails to build, the system defaults to a base image and surfaces clear warnings. This allows the agent to attempt a recovery or continue working rather than failing the entire process immediately.

For enterprise teams bypassing manual Dockerfile creation, Cursor introduced an agent-led setup tool, currently in private beta. This tool inspects a repository and automatically generates the required environment configuration based on the codebase’s dependencies.

Governance and Audit Logging

As agents gain access to broader infrastructure, security controls become the limiting factor for deployment. Cursor has introduced scoped egress, restricting network access and secret exposure to specific development environments rather than the entire workspace.

Every workspace maintains a strict version history. Administrators can audit every action taken by team members or agents within these environments and trigger rollbacks if necessary. A recent partnership with Graphite allows engineering teams to review and initiate agent-generated pull requests directly within their existing code review platforms.

If you build distributed systems, this release shifts how you assign tasks to programmatic agents. You no longer need to manually synchronize changes across repo boundaries for the AI. You can provide the agent with a multi-repo workspace, define its dependencies as code, and scope its credentials strictly to the required package registries.