Show HN: Ramet – Persistent, forkable workspaces for AI agents

Category: infrastructure

Tags: ai-agents, workspace-management, persistence, sandboxing, state-management

Score: 7.3/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 6)

Ramet provides persistent, forkable workspaces for AI agents, enabling save/restore of agent state across sessions and sandbox replacements. It offers isolated writable workspaces from a shared immutable base, supports selective promotion of state, and integrates with various agent frameworks. The project addresses the challenge of managing filesystem state in agentic workflows, particularly for parallel attempts and reproducibility, with a focus on efficiency and flexibility.

Target audience: AI developers, ML engineers, DevOps

Repository: https://github.com/shenli/ramet · Go · Apache-2.0 · 6 stars

View on Hacker News