AgentPool — Cloud agent workspace snapshot & reconnect tool
For developers using ephemeral cloud agent VMs (Muse, Railway), turn their reclaimable persistent dir into a publicly reachable, recoverable work VPS
- Problem
- Platforms like Meta Muse give each user a Linux VM, but it has no public IP and auto-shuts down when idle; the root is reset while only /home/hatch/workspace persists. Users who want a VPS must manually configure tunneling, keep-alive, and move tools into the persistent directory. Railway's ssh railway.new has similar ephemeral instance and quota limits. Nothing automates this today.
- Solution
- A minimal CLI + lightweight daemon: detects agent VM environments, puts the user's toolchain and configs into the persistent directory, establishes a secure public entry via built-in tunneling (cloudflared/frp), and sends periodic keep-alive requests via cron/webhook to prevent reclaim. One command turns an ephemeral VM into a remotely SSH-able, long-running dev/crawler/short-task VPS.
- Users
- Indie devs, AI agent hobbyists, students, and script writers who need temporary cloud dev environments without buying a VPS
- MVP
- Detect whether current environment is a Muse/Railway agent VM; install user-specified toolchain (Node/Python/scripts) inside the persistent directory; start built-in tunnel with one command and generate a random public URL; provide a daemon for periodic activity to prevent shutdown; include restore command to recover config after instance restart.
- Revenue
- Open-source core + paid cloud-hosted tunnel/keep-alive scheduler (subscription $5-9/mo); or a one-time buyout desktop one-click setup tool.
- Why now
- Free/cheap ephemeral agent VMs from Muse, Railway, etc. are being widely used as lightweight VPS by developers, but platform limits and manual setup are painful; an indie dev can build a more stable, worry-free tool quickly without relying on big platform APIs.