AI Deck

exe.dev — Persistent Linux VMs You Reach Over SSH, With an AI Coding Agent Living Inside

exe.dev is a cloud development platform that gives you persistent Linux VMs you can use simply by connecting over SSH. Its defining trait is a pooled resource model: you buy CPU, RAM, and disk as a single pool and spin up as many VMs as you like within it, so throwaway experiment boxes and long-running servers coexist under the same bill without per-machine contracts. Every VM ships with Shelley, a web-based AI coding agent, so you can hand off work straight from a browser or a phone. It is built by Bold Software, Inc. — led by David Crawshaw, a Tailscale co-founder — and Shelley is positioned as the successor to their earlier coding agent, Sketch.

Key Features

  • SSH is the management interface: Creating, connecting to, and managing VMs all happens over SSH. There is no separate dashboard to log into; you work with your environments the way you would with ssh exe.dev
  • Pooled resource purchasing: You buy vCPU, RAM, and disk as a total allowance and create as many VMs as fit inside it. Because billing is framed around resources rather than machine count, short-lived sandboxes are easy to justify
  • Real Linux with root access: Each VM is an ordinary Linux box you enter as root, with apt and systemd available. There are fewer container-style constraints, so existing server operations knowledge carries over
  • Automatic HTTPS and DNS: Web servers you start on a VM automatically receive an HTTPS URL with a certificate. No hand-rolled tunnels or reverse proxies required
  • IAM-based sharing controls: VMs are private by default and can be shared individually with teammates or outside collaborators through IAM settings. Temporary demo access works the same way
  • Shelley, the built-in AI agent: A web-based coding agent that is multi-conversation, multi-model, and multi-modal. It keeps conversation history in a SQLite database and supports diff review with the Monaco editor, screenshot analysis, and proxying to the web server you are developing. Because the VM persists, you can shape the environment first and then let the agent work in it

Pricing

PlanPriceResourcesWhat’s included
Personal$20/month2 vCPU / 8 GB RAM (shared across all VMs)Up to 50 VMs, 100 GB pooled disk, 200 GB data transfer, monthly Shelley Token allocation
Team$25/user/month2 vCPU / 8 GB RAM per userUp to 50 VMs per user, 100 GB pooled disk, 250 GB data transfer, admin and user management, SSO, burstable capacity sharing among teammates
Reserved Cloud Pool$35.84/hour512 vCPU / 2048 GB RAMThousands of VMs, admin and user management, SSO, AWS VPC integration

Overages are billed at $0.08 per GB of disk per month and $0.05 per GB of data transfer. Each VM is allocated 25 GB of disk by default, adjustable via CLI arguments. The official pricing page lists no free tier.

Pricing is current as of August 2026. Check the official site for the latest rates.

Pros & Cons

Pros

  • Everything runs through SSH, so there is little to learn and it slots into existing workflows
  • The pooled model means adding VMs does not add contracts, making “just spin one up” a low-friction move
  • Environments persist, so you can install the tools you need before handing work to an AI agent
  • HTTPS URLs are provisioned automatically, shortening the path to a working demo or a shared preview
  • Root access to ordinary Linux lets you bring your existing server operations know-how as-is

⚠️ Cons

  • There is no free tier, so evaluation requires a paid plan from the start
  • The Personal plan’s 2 vCPU / 8 GB RAM is shared across all VMs, which can feel tight for heavy builds or highly concurrent workloads
  • SSH and command-line use are assumed, which raises the bar for people who work mostly in GUIs
  • Shelley usage is governed by a monthly token allocation, so heavy users need to keep an effective ceiling in mind
  • It is a relatively new service, and its ecosystem and body of case studies are thinner than those of established clouds

Comparison with Similar Services

Criteriaexe.devGitHub CodespacesGitpodFly.io
Primary usePersistent VMs plus an AI agent runtimeCloud dev environment (IDE-centric)Cloud dev environment (IDE-centric)Application deployment platform
Access methodSSH (management included)Browser / VS CodeBrowser / IDE integrationCLI (flyctl) plus deploys
PrivilegesRoot on a Linux VMContainer userContainer userContainers / micro VMs
AI agentShelley built inPairs with GitHub CopilotBring your own agentNone built in
Billing modelPooled resource purchase (flat rate)Usage hours plus storageUsage hoursResource-based usage

Who Is It For

  • People who want to hand an entire runtime to an AI coding agent but find rebuilt-every-time throwaway environments limiting
  • Developers who frequently stand up and tear down test servers and want to avoid the contract and billing math each time
  • Developers who want to experiment on real Linux with root access without cluttering their local machine
  • Anyone who wants an HTTPS URL on what they built so they can show it to a team or client right away
  • People comfortable with SSH and the command line who do not need a GUI dashboard

Summary

exe.dev turns the idea of “a cloud for developers and their agents” into something concrete: persistent VMs you reach with a single SSH command, paid for as a pool of resources. What separates it from traditional throwaway sandboxes is that you can hand an agent a whole runtime, and that the runtime survives long enough to be shaped over time. With no free tier, the sensible way in is to start on the Personal plan, spin up a few small VMs, give Shelley one real task, and see whether the model fits how you work.

← Blog