Policy & Community

Eclipse Foundation releases Enclave 1.0, open-source sandbox for autonomous AI coding agents

Key Takeaway

Eclipse Enclave 1.0 provides a vendor-neutral, open-source runtime that isolates AI coding agents in policy-controlled containers to enable autonomous operation without unrestricted system access.

Summary

The Eclipse Foundation has released Eclipse Enclave 1.0, an open-source runtime and sandbox designed to let AI coding agents operate autonomously within isolated, policy-controlled environments. The project supports Claude Code, Codex, OpenCode, and other agents through an extensible architecture, allowing developers to run multiple agents in parallel on separate tasks without granting them full access to their development systems.

Enclave originated at EclipseSource, which contributed the sandboxing technology to the Eclipse Foundation to broaden community access and collaborative development. Each agent session runs in an isolated container with defined access to files, networks, and development resources, addressing security and governance concerns that arise as coding agents move beyond suggestions to independently executing commands, installing dependencies, and modifying software.

Architecture and security model of Eclipse Enclave

  • Container-based isolation: Each agent session executes in an isolated container with configurable, defined access to files, networks, and development resources rather than unrestricted host system access.
  • Multi-agent parallel execution: Enclave supports running multiple agents simultaneously on separate tasks, with isolation between sessions to prevent cross-contamination and simplify review.
  • Credential protection mechanisms: The runtime includes controls for restricting access to external services and protecting API keys and OAuth tokens used for third-party service authentication.
  • Persistent session state: Agents can resume previous work without losing progress or configuration, enabling longer-running autonomous tasks across supported coding agents.
  • Extensible agent support: The architecture currently integrates Claude Code, Codex, and OpenCode, with hooks for additional tools to be added by the community or vendors.

Why it matters

Eclipse Enclave addresses a critical gap in the rapidly expanding AI-assisted development toolchain by providing a vendor-neutral alternative to proprietary sandboxing solutions, reducing lock-in risks for enterprises adopting multiple coding agents. For Linux and open-source developers, its integration with the broader Eclipse ecosystem including Theia IDE and Open VSX Registry creates a coherent open-source alternative to GitHub Copilot's closed environment. The project's placement under the Eclipse Foundation's governance, alongside the Sovereign AI Foundation initiative, signals growing institutional investment in open-source AI infrastructure that prioritizes developer autonomy and organizational control over cloud-dependent, vendor-controlled alternatives.

What to do

  • If you use AI coding agents like Claude Code, Codex, or OpenCode, consider running them in Eclipse Enclave 1.0 to isolate each session in a policy-controlled container.
  • Restrict agent access to files, networks, and external services using Enclave's built-in controls, and protect API keys and OAuth tokens through its credential protection mechanisms.
  • Run multiple agents in parallel on separate tasks to increase productivity while keeping their work isolated and easier to review.
  • Resume previous agent sessions without losing progress or configuration, enabling longer-running autonomous tasks.

Original reference: newsroom.eclipse.org →