ephemeral-ai-lab/layerfs is a Rust project that helps multiple agents work in their own writable filesystem workspaces based on a shared commit history. The latest release identifiable from the provided materials is v1.1.0. However, the materials do not specify the concrete changes in this version, so no claim can be made about what changed.
In this model, agents work in parallel without copying the base, and store their changes as deduplicated checkpoints. They can also branch from the working history, roll back to a previous state, and merge. In other words, it is a way to manage shared state across agents while preserving write access in each workspace. The README describes ephemeral execution environments and collaboration among distributed agents as its primary use cases, and points to the DeepSeek Harness plugin as the current usage path. The API is divided into agents, backends, and runtimes. The repository is licensed under the MIT License.
However, the README explicitly states that its performance comparisons are measurements from an earlier version, so they cannot be treated as evidence of v1.1.0's performance. The README also warns against using LayerFS as a standalone data storage system. Therefore, while the materials do confirm the approach of managing agent workspaces on top of a shared history, the performance of the latest version and its suitability as a standalone repository are not supported by these materials alone.