Competitive positioning: Linux boxes you SSH into
This page covers only alternatives that give someone a persistent Linux machine they SSH into. Every competitor price below comes from that vendor’s own public page or API, checked 2026-10-23 and linked in the row. Where a number couldn’t be checked, the page says so instead of guessing. Totals marked derived are arithmetic on those published rates.
The exe.dev angle in one breath
One flat price for a pool of boxes, and every box already has HTTPS, login and sharing.
- One budget, many machines. Your VMs share CPU and RAM, and you pay for the underlying resources, not per VM (https://exe.dev/docs/what-is-exe). Personal 2 vCPU is $15/month for up to 50 VMs sharing 2 vCPU and 4 GB (https://exe.dev/pricing).
- HTTPS on every box. Each VM is served at
https://vmname.exe.xyz/with TLS handled by exe.dev’s proxy (https://exe.dev/docs/proxy). - Login built in. Sites are private by default, and your app receives the visitor’s identity in
X-ExeDev-Emailheaders (https://exe.dev/docs/login-with-exe). - Sharing like a document. exe.dev says you can share a web server “as easily as you can share a Google Doc” (https://exe.dev/docs/what-is-exe). In practice that’s
share add <vm> <email>, or--rootfor shell access (https://exe.dev/docs/sharing). - Agents and integrations included. Shelley runs on every default-image VM, and integrations keep credentials off the VM (https://exe.dev/docs/shelley/intro, https://exe.dev/docs/integrations).
Be honest about price per dedicated vCPU. On raw compute, budget VPS hosts are cheaper. exe.dev wins when many boxes sit idle most of the time and share a pool, and when the customer counts the time it takes to set up TLS, auth and sharing on every VPS.
Price reference: one 2 vCPU / 4 GB box, always on
| Provider and plan | What you get | List price | Source |
|---|---|---|---|
| exe.dev Personal 2 vCPU | Up to 50 VMs sharing 2 vCPU / 4 GB, 100 GB disk, 200 GB egress | $15/mo | https://exe.dev/pricing |
| exe.dev standalone VM | 2 vCPU / 4 GiB of its own, per second | $0.105/h, about $76.70/mo | https://exe.dev/docs/standalone-vms |
| Hetzner Cloud CX23 (Cost-Optimized) | 2 vCPU, 4 GB, 40 GB NVMe, 20 TB traffic (EU), IPv4 included | from €5.99/mo excl. VAT (€0.0096/h). Listed as “not available” when checked | https://www.hetzner.com/cloud/cost-optimized |
| DigitalOcean v5 Droplet (shared vCPU) | Built per resource: 2 shared vCPU, 4 GiB, 80 GiB boot disk | $0.0587682/h, about $42.93/mo derived | https://docs.digitalocean.com/products/droplets/details/pricing/ |
| Linode/Akamai Linode 4GB | 2 vCPU, 4 GB, 80 GB disk, 4 TB transfer | $24/mo ($0.036/h) | https://api.linode.com/v4/linode/types (Akamai’s public price API) |
| Vultr Cloud Compute, Regular Performance | 2 vCPU, 4 GB, 80 GB, 3 TB bandwidth | $20/mo ($0.03/h) | https://www.vultr.com/pricing/ |
| Vultr Cloud Compute, High Performance | 2 vCPU, 4 GB, 100 GB NVMe, 5 TB bandwidth | $24/mo ($0.036/h) | https://www.vultr.com/pricing/ |
| AWS Lightsail, Linux, with public IPv4 | 2 vCPU, 4 GB, 80 GB SSD, 4 TB transfer | $24/mo | https://aws.amazon.com/lightsail/pricing/ |
| Fly.io Machine shared-cpu-2x 2 GB, plus 2 GB RAM and an 80 GB volume | 2 shared vCPU, 4 GB, 80 GB volume | $13.39 + $12.00 + $12.00 = $37.39 per 30 days derived | https://fly.io/docs/about/pricing/ |
| Fly.io Machine performance-2x 4 GB | 2 performance vCPU, 4 GB (no volume) | $66.00 per 30 days | https://fly.io/docs/about/pricing/ |
| GitHub Codespaces 2-core | Billed per active hour, plus storage | $0.18/h; 176 h (8 h × 22 days) = $31.68 derived | https://docs.github.com/en/billing/concepts/product-billing/github-codespaces |
| Daytona sandbox, 2 vCPU / 4 GiB | Per second while running | $0.1656/h derived ($0.0504/vCPU-h + $0.0162/GiB-h) | https://www.daytona.io/pricing |
| Coder Community | Self-hosted, you supply the machines | Free software; Premium priced “annually per user”, no public number | https://coder.com/pricing |
| Gitpod / Ona | Ephemeral environments in Ona Cloud or your VPC | No public price; new signups closed | https://ona.com/docs/ona/getting-started |
Ten boxes, derived
Say a team wants ten 2 vCPU / 4 GB boxes, each idle most of the day.
| Option | Monthly derived | Notes |
|---|---|---|
| 10 × Hetzner CX23 | €59.90 excl. VAT | Cheapest raw compute, if stock is available. You add TLS, auth and sharing yourself |
| 10 × Vultr Regular Performance | $200 | Same |
| 10 × Linode 4GB, or 10 × Lightsail | $240 | Same |
| 10 × Fly shared-cpu-2x setup above | $373.90 | Less if Machines auto-stop when idle |
| exe.dev Work, 4 vCPU pool shared by all ten | $75 + 1 unit × $0.105 × 730.5 h = $151.70 | HTTPS, login, sharing and Shelley on every box. 200 GB disk included for the pool |
| exe.dev Work, 20 vCPU pool (no sharing) | $75 + 9 units × $0.105 × 730.5 h = $765.32 | If every box truly needs its own 2 vCPU all month, exe.dev costs more |
The selling point is the shared pool. If the customer’s boxes are busy all the time, lead with the built-in HTTPS, auth and sharing, not with price.
Closest match: Fly.io Machines
What they are. Fast-launching VMs billed per machine, “by the second while they’re running”. Stopped Machines bill only for rootfs at $0.15/GB/month, volumes are $0.15/GB/month, and a dedicated IPv4 is $2/month. There’s no free tier for new organizations, only a trial (https://fly.io/docs/about/pricing/).
Where exe.dev wins
- One pool, many boxes. Fly bills each Machine. exe.dev bills a pool that any number of VMs share (up to 1,000 per Work pool, https://exe.dev/docs/billing/work).
- Disk included. Personal 2 vCPU includes 100 GB. On Fly, 100 GB of volume is $15/month derived.
- Private by default with login. Fly gives you networking. exe.dev gives you
X-ExeDev-Emailandshare addwith no auth code to write. - A box you live in.
ssh vmname.exe.xyz, a real disk, systemd, Docker and an agent already there (https://exe.dev/docs/faq/how-exedev-works, https://exe.dev/docs/faq/docker).
Where Fly wins. Anycast IPs with a free shared IPv4 per app, many regions, pricing that drops when Machines auto-stop, and a 40% discount on reserved compute blocks.
Objection: “Fly already gives us micro-VMs for a few dollars each.” Answer: “For one app, sure. For thirty dev and agent boxes, you’d pay for each one and then build login and sharing yourself. On exe.dev those thirty boxes share one pool, and each one has HTTPS, login and sharing from the first second.”
Raw VMs, billed per VM: Hetzner, DigitalOcean, Linode/Akamai, Vultr, AWS Lightsail
What they are. Classic VPS hosts. You get a VM with a public IP, then you set up DNS, TLS, a reverse proxy, auth and backups.
- Hetzner bills servers whether they’re on or off, “until they are deleted” (https://www.hetzner.com/cloud/cost-optimized).
- DigitalOcean bundled-plan Droplets are billed when powered off too. Bundled plans cap usage at 672 hours a month, while a continuously running v5 Droplet’s total varies with the hours in the month (https://docs.digitalocean.com/products/droplets/details/pricing/).
- Linode keeps billing a powered-off VM (https://www.linode.com/pricing/).
- Vultr charges 20% more for automatic backups and $0.01/GB over quota (https://www.vultr.com/pricing/).
- Lightsail bundles a static IP, DNS management and browser SSH into the price (https://aws.amazon.com/lightsail/pricing/).
Where exe.dev wins
- No per-box tax. No IPv4 per VM, no certificate per VM, no nginx config per VM. exe.dev pools IP addresses because paying $2/month per IPv4 for every VM would be “far more than is reasonable”, and terminates TLS at its proxy instead (https://exe.dev/docs/architecture).
- About two-second boots from a container image onto a block device (https://exe.dev/docs/faq/how-exedev-works).
- Sharing and login for internal tools, with no auth code.
Where they win. Lowest price per dedicated vCPU, a public IP per server, any port, kernel choice and private networks. exe.dev says “you don’t get to choose which kernel you’re using”, and there’s no private network between VMs (https://exe.dev/docs/faq/how-exedev-works, https://exe.dev/docs/faq/cross-vm-networking).
Objection: “A Hetzner box is €5.99. Why pay more?” Answer: “For one always-busy server, Hetzner is great. For a team’s worth of dev boxes, previews and agent machines, count the hours spent on certs, auth and sharing per box. exe.dev puts ten boxes in a $151.70 pool with all of that done.”
Repo-tied dev environments: GitHub Codespaces, Gitpod/Ona
What they are.
- Codespaces bills per active hour, from $0.18/h for 2-core to $2.88/h for 32-core, plus $0.07/GB-month of storage. Personal accounts include 120 hrs (Free) or 180 hrs (Pro) of compute a month, counted with a multiplier equal to the core count, and organization plans include no free quota (https://docs.github.com/en/billing/concepts/product-billing/github-codespaces). A codespace stops after 30 idle minutes by default (settable from 5 to 240 minutes) and has a maximum lifetime of 12 hours (https://docs.github.com/en/codespaces/setting-your-user-preferences/setting-your-timeout-period-for-github-codespaces).
- Gitpod now redirects to Ona (gitpod.io → ona.com). Ona describes “ephemeral, isolated environments”. On the Core plan, stopped environments are archived after 3 days and deleted 4 days later. New customer signup is closed (https://ona.com/docs/ona/getting-started, https://ona.com/docs/ona/environments/archive-auto-delete).
Where exe.dev wins
- Boxes that stay up. A VM keeps running “when we close our laptop lid” (https://exe.dev/docs/why-exe). A 2-core codespace running all month would be $131.49 derived, but a codespace has a 12-hour maximum lifetime anyway.
- Not tied to one repo. An exe.dev VM is a general Linux box: run a service, a database or an agent.
- A URL you can share. Every box has HTTPS, login and
share add.
Where they win. One click from a GitHub repo, devcontainer.json, and prebuilds. Codespaces gives individuals a free monthly quota.
Objection: “We already have Codespaces through GitHub.” Answer: “Keep it for quick PR edits. When the work has to keep running, like an agent overnight, a preview a client can open, or an internal tool, put it on a box that doesn’t stop after 30 idle minutes or 12 hours and already has HTTPS and sharing.”
Agent sandboxes you can SSH into: Daytona
What it is. Per-second sandboxes priced at $0.0504/vCPU-h, $0.0162/GiB-h and $0.000108/GiB-h of storage after 5 free GiB, with $200 of free compute (https://www.daytona.io/pricing). Sandboxes are Linux containers by default, with VM sandboxes available. Defaults are 1 vCPU, 1 GB and 3 GiB of disk, and organizations get a maximum of 4 vCPU, 8 GB and 10 GB per sandbox. Auto-stop defaults to 15 minutes. SSH access is supported (https://www.daytona.io/docs/en/sandboxes.md, https://www.daytona.io/docs/llms.txt).
Where exe.dev wins
- Real VMs by default, isolated from each other by KVM, with persistent disks (https://exe.dev/docs/enterprise-single-tenant).
- Secrets stay off the box. Integrations inject credentials at the network edge, and
egress <vm> integrationslocks outbound traffic (https://exe.dev/docs/integrations). - Bigger boxes and flat pricing. Work pools reach 512 vCPU. A 2 vCPU / 4 GiB box left on all month is about $120.97 on Daytona derived, versus a share of a flat pool.
Where Daytona wins. Sandboxes that spin up “in milliseconds”, GPU options, and Windows sandboxes (https://www.daytona.io/pricing).
Objection: “Our agents need sandboxes, not VMs.” Answer: “exe.dev standalone VMs are sandboxes too: billed per second, with capacity of their own (https://exe.dev/docs/standalone-vms). The difference is that the agent gets a real VM, and your tokens never land on it.”
Self-hosted: Coder
What it is. Coder describes itself as “governed, self-hosted environments for your developers and the coding agents you approve”. Community is free and open source. Premium is priced “annually per user” with no public number (https://coder.com/pricing).
Where exe.dev wins. Nothing to run. With Coder you still buy, patch and network the machines underneath, and then run Coder on top.
Where Coder wins. Everything stays on infrastructure the customer already owns, in any cloud or air-gapped. For exe.dev’s answer to that, see single-tenant hosts and BYOC (https://exe.dev/docs/enterprise-single-tenant, https://exe.dev/docs/enterprise-byoc).
Objection: “We’d rather self-host Coder.” Answer: “Then you’re running a platform team for dev boxes. exe.dev Enterprise can run on dedicated hosts or in your own cloud account, without you operating the control plane.”
DIY: your own box plus Tailscale Funnel or Cloudflare Tunnel
What it is.
- Tailscale Funnel routes internet traffic to a service on a tailnet device. It’s in beta, available on all plans, listens only on ports 443, 8443 and 10000, and has “non-configurable bandwidth limits” (https://tailscale.com/kb/1223/funnel). Tailscale Personal is $0 for up to 6 users (non-commercial use), Standard $8 and Premium $18 per user per month (https://tailscale.com/pricing).
- Cloudflare Tunnel runs
cloudflared, which “creates outbound-only connections to Cloudflare’s global network”, so the box needs no public IP (https://developers.cloudflare.com/cloudflare-one/networks/connectors/cloudflare-tunnel/).
Where exe.dev wins. DIY means one machine, one tunnel and one auth setup at a time, on hardware you maintain. exe.dev gives every new box HTTPS, login and sharing in about two seconds, inside one pool.
Where DIY wins. Free if you already own the hardware, local GPUs, and full control.
Objection: “I can do this with a spare box and Cloudflare Tunnel.” Answer: “For one service, yes. Now do it for every branch preview and every agent, with sharing by email and no machine under your desk.”
Honest limits to bring up first
These come from exe.dev’s docs. Raising them early builds trust.
| Limit | Source |
|---|---|
| No kernel choice | https://exe.dev/docs/faq/how-exedev-works |
| No private network between VMs | https://exe.dev/docs/faq/cross-vm-networking |
| No public IP per VM; one public HTTPS port | https://exe.dev/docs/proxy |
/exec has a 30 s timeout, 64 KB body, no stdin/pty | https://exe.dev/docs/https-api |
Receiving email works only on *.exe.xyz; “No custom domains yet” | https://exe.dev/docs/receive-email |
| PDX is closed to new accounts | https://exe.dev/docs/regions |
| Pools above 32 vCPU need approval | https://exe.dev/docs/pools |
| Nested VMs are Enterprise only | https://exe.dev/pricing |
| Cloudflare’s proxied (orange cloud) mode breaks custom domains | https://exe.dev/docs/cnames |
| RPO/RTO for the streaming disk copy aren’t published | https://exe.dev/docs/architecture [verify] |
Lock-in objection
“As long as it runs on Linux, exe.dev supports it … If you need to migrate off to another provider, rsync will get you pretty far.” The default image’s Dockerfile is public (https://exe.dev/docs/lockin).