25 days ago
- AI coding agents need autonomy to run commands, install packages, and edit files efficiently, but granting this directly on host machines poses risks like system damage, secret leaks, and interference.
- Eclipse Enclave provides a sandboxed runtime with isolated containers, filesystems, process trees, and network stacks, allowing multiple agents to work in parallel without interference via separate Git worktrees.
- The platform includes a sidecar gateway to restrict and log outbound network traffic, along with features for auth, configuration, history persistence, and a control center for managing agents.
- Its isolation, logging, and policy infrastructure offers audit trails, network access logs, and dependency records, helping organizations comply with regulations like the EU AI Act and Cyber Resilience Act.
- Eclipse Enclave separates agent execution from identity, treating agents as pluggable workloads to enable community focus on isolation, observability, and governance, while allowing users to mix agents and editors.