Skip to content
Shotcup

Shotcup and other sandboxes

MicroVMs and containers give an agent a whole computer. Shotcup gives it a fresh JavaScript engine per run, which is less, on purpose: short programs start fast, run the same way twice and cost a run, not a machine. Here is what each option is, from the vendors' own pages, with dates.

Side by side

Shotcup, E2B, Daytona, Vercel Sandbox and Cloudflare compared
QuestionShotcupa fresh JavaScript engine per runE2BFirecracker microVMs with full LinuxDaytonaContainers and VMs from images or snapshotsVercel SandboxFirecracker microVMs with UbuntuCloudflareWorkers isolates, Dynamic Workers and Linux containers
What runs
  • A fresh GocciaScript engine per run in a pooled worker process, with a virtual filesystem. JavaScript only: no operating system, no npm at run time.
Start-up
  • Measured by us: 89.5 ms p50 server time for a whole run on shotcup.dev (2026-10-08), sandbox included.
  • Vendor: each session gets a Linux machine that boots from a snapshot; the site gives no create-time figure.e2b.dev,
How usage is billed
  • A monthly allowance of runs and compute-hours per plan; past it, opt-in overage of $0.05 per 1,000 runs and $0.08 per compute-hour under a spend cap.
  • $0.0504 per vCPU-hour and $0.0162 per GiB-hour, billed per second; minimum 1 vCPU and 1 GiB.daytona.io,
  • Active CPU $0.128 per hour, memory $0.0212 per GB-hour, billed "in 1 minute minimum increments"; $0.60 per million creations; minimum 1 vCPU and 2 GB.vercel.com,
Free tier and entry plan
  • Free: 30,000 runs and 3 h of compute a month. Pro: $4.99 a month for 100,000 runs and 10 h.
  • Hobby: $0 with a one-time $100 credit. Pro: "$150 monthly plan fee, plus usage".e2b.dev,
  • Hobby includes 5 CPU-hours, 420 GB-hours and 5,000 creations; Pro has a $20 platform fee.vercel.com,
Concurrency and run length
  • 1 concurrent run on Free, 10 on Pro, up to 100 on Platform. A run lasts at most 5 s on Free and 5 min on Scale.
  • 10 concurrent on Hobby, 10,000 on Pro; sessions of 45 minutes on Hobby and 24 hours on Pro.vercel.com,
State between runs
  • Disposable by default; saved sandboxes keep their files between runs. Seeded clocks and randomness make runs repeatable, and a run that needs approval pauses and resumes.
MCP and agents
  • A hosted MCP server and a REST API on one key; a TypeScript SDK, an AI SDK sandbox and an eve provider.
  • Integrations with the AI SDK, OpenAI and Anthropic SDKs and others; no sandbox MCP server.vercel.com,

Start-up figures for the other vendors are their own claims; we did not measure them. Shotcup's is a measurement, explained on the performance page. Prices change: follow the links for the current ones.

What 1,000 short runs cost

Isolates billed by CPU time are the cheapest per run, and Shotcup is not: at its overage rates a run costs more than a Worker. What a plan buys is a fixed monthly price with the runs included, the sandbox around each run, and the limits, artifacts and MCP server on top. MicroVMs billed by the minute cost far more for work this short.

Usage cost of 1,000 short runs, in US dollars, computed by Shotcup
Sandbox1,000 runsBilled as
Shotcup$0.051overage rates; included in a plan's allowance
Cloudflare Workers$0.0013per-request and CPU-ms rates
Cloudflare Containers$0.0043 s of life at the smallest size
E2B$0.015about 1 s per sandbox
Daytona$0.0191 vCPU and 1 GiB minimum
Vercel Sandbox$0.711-minute minimum, 1 vCPU and 2 GB
Cloudflare Dynamic Workers$2.00a unique worker for every run

Computed by us from each vendor's published rates on 8 October 2026, not measured: 1,000 runs, each a fresh sandbox doing 50 ms of CPU with 64 MB; each vendor billed at its minimum size and increment, or about one second of life including boot; base plan fees left out.

One at a time