Your GitHub Actions, up to 3× faster
Same workflows, faster machines. One setup pull request — and if we can't take a job, it runs on your runners, so CI never breaks.
macOS M4 Pro from $0.028/min (GitHub: $0.062) · Linux 8 vCPU $0.009/min (GitHub: $0.022) · credit from $20, never expires
- CI jobs run on compiler.dev
- 3,000+
- Repositories that ran jobs here
- 28
- Common stacks benchmarked
- 19 · 15 faster
- Seeded bugs caught by code review
- 6 of 6
Stacks like yours, measured: Next.js + Postgres 1.3× · Go 2.3× · Spring Boot 1.6–3.0× · Rails, Laravel, .NET, Phoenix, Rust and more — see your row below, then measure your own jobs before switching anything.
Code in a box New
Your own Linux machine for coding agents
Bring your Claude Code or Codex key. Open it from any browser. It stops when you're idle and costs about $0.39 an hour while it runs.
Agents use fewer tokens here: terminal output is compacted before it reaches the model (RTK), typically 60–90% less on builds, tests and git.
- 4 vCPU, 16 GB, 100 GB disk that keeps your repos and work
- Up to 3 agents at once, each in its own worktree
- Watch live output and the diff; open the PR when it's right
- A one-time link opens the same chat on your phone
While it runs
$0.39 / hour
Disk: 100 GB at $10 a month, kept while it's stopped.
From your prepaid credit. Model usage is billed by Anthropic or OpenAI on your own key.
Start coding in your box How it worksMeasured on common stacks
Full CI runs, 10 test shards, real database · 8 October 2026
| Stack | GitHub-hosted | compiler.dev | Speed | Cost per run |
|---|---|---|---|---|
| Angular + Spring Boot (Gradle) + Postgres (JUnit + Playwright) | 6m 57s | 2m 18s | 3.0× faster | $0.18 → $0.14 (−20%) |
| Remix + Drizzle + Postgres (Vitest + Playwright) | 3m 07s | 1m 18s | 2.4× faster | $0.19 → $0.13 (−34%) |
| React + Go + Postgres (go test -race) | 6m 13s | 2m 42s | 2.3× faster | $0.37 → $0.28 (−24%) |
| React + ASP.NET Core + SQL Server (xUnit) | 3m 34s | 1m 41s | 2.1× faster | $0.17 → $0.20 (+18%) |
| Phoenix LiveView + Postgres (ExUnit) | 2m 07s | 1m 03s | 2.0× faster | $0.13 → $0.10 (−25%) |
| Turborepo + Next.js + tRPC + MySQL (Vitest + Playwright) | 3m 29s | 1m 53s | 1.9× faster | $0.17 → $0.17 (−5%) |
| Next.js + Go + Kafka + Redis (docker compose) | 1m 16s | 0m 44s | 1.7× faster | $0.13 → $0.11 (−14%) |
| SvelteKit + FastAPI + Postgres + Redis (pytest + Vitest) | 3m 46s | 2m 18s | 1.6× faster | $0.21 → $0.22 (+3%) |
| Spring Boot (Maven) + Kafka + Postgres (Testcontainers) | 1m 54s | 1m 12s | 1.6× faster | $0.12 → $0.11 (−12%) |
| React + Rust (Axum) + Postgres (cargo test) | 2m 37s | 1m 45s | 1.5× faster | $0.16 → $0.14 (−11%) |
| Next.js + Supabase (Playwright) | 4m 03s | 2m 48s | 1.4× faster | $0.26 → $0.23 (−15%) |
| Nx: Angular + NestJS + Postgres (Jest + Playwright) | 3m 17s | 2m 26s | 1.4× faster | $0.25 → $0.23 (−9%) |
| Nuxt + Laravel + MySQL (Pest) | 2m 37s | 1m 57s | 1.3× faster | $0.13 → $0.15 (+16%) |
| Next.js + Express + Postgres (Vitest + Playwright) | 2m 31s | 2m 00s | 1.3× faster | $0.19 → $0.19 (−2%) |
| React (Vite) + NestJS + Postgres (Cypress) | 3m 23s | 3m 06s | 1.1× faster | $0.19 → $0.16 (−16%) |
First job queued to last job done; GitHub ubuntu-latest (2-core, $0.006/min) vs compiler.dev (8 vCPU, $0.009/min). 1 GitHub run and 1–2 compiler.dev runs per stack. Of 19 stacks, 15 are faster (above), 2 about the same, 1 slower (single-threaded pytest), 1 not yet compared. Comparison mode measures your own jobs before you switch anything.
- Automatic fallback. Down, slow or full? The job runs on your runners.
- Deploy jobs never move. Only Linux build and test jobs.
- Undo any time. Set 0% or revert the setup PR.
Pricing
About half what GitHub charges for the same machine: 46–55% less on macOS and 59% less on 8 vCPU Linux. Buy credit up front (from $20, never expires); jobs draw it down per minute. At $0 your jobs simply run on your own runners — CI never breaks.
| Machine | GitHub-hosted | compiler.dev | Availability |
|---|---|---|---|
| macOS · Apple M4 Pro | |||
| Dedicated host, reserved for you | ≈ $4,400/month (5-core M2 Pro, busy all day) | from $3,595/month −18% | Talk to us |
| Full 14-core, fresh VM per job | $0.102/min (5-core M2 Pro) | $0.055/min −46% | Early access |
| 7-core VM | $0.062/min (3-core M1) | $0.028/min −55% | Early access |
| Linux x64 | |||
| 8 vCPU / 16 GB (default) | $0.022/min (8-core) | $0.009/min −59%under 1¢ a minute | Live |
| 2 / 4 / 16 / 32 vCPU | $0.006 / $0.012 / $0.042 / $0.082 | $0.0035 / $0.0065 / $0.018 / $0.036 42–57% less | On request |
| Code in a box · your own machine for coding agents | |||
| 4 vCPU / 16 GB, 100 GB disk (what's this?) | — | $0.0065/min ≈ $0.39/hour while running · disk $10/month | Live |
Enterprise · Done for you
Contact us
- We move your CI for you and tune it: test splitting, caches, Mac images with your Xcode versions
- Faster than your current CI, or you don't pay for the migration
- Dedicated Apple M4 Pro hosts for your team, unlimited minutes
- A named engineer, a shared Slack channel, a monthly CI speed report
- SLA, SSO, committed Linux spend, code review included
CI + Review Early access
$25 / active developer / month
- Everything in CI and Code review
- Includes $20 of runner-minutes per developer each month (about 2,200 minutes), then $0.009/min
- Billed monthly in arrears on actual active developers — no prorating, no annual lock-in
- Spending cap on runner usage, same as pay as you go
Code review
$15 / active developer / month
- Reviews every pull request for bugs, security and regressions — not style your linter already catches
- Explains failing CI on our runners next to the line that caused it
- A seat is someone who opened or pushed to a PR that month; bots never count
- Paid from your prepaid credit, once per developer per month, no prorating; optional monthly cap
- Turn it on for all repositories or only the ones you pick
CI · Prepaid
$0.009 / runner-minute
- Prepaid credit from $20 — unused credit never expires; optional auto top-up
- No seats, no minimums; monthly spending cap — at the cap, jobs fall back to your runners
- 8 vCPU / 16 GB Linux machines, fresh for every job — GitHub charges $0.022/min for its 8-core runner
- Pay by card via Stripe; receipt by email
Linux is measured per job at the per-minute rate; sizes other than 8 vCPU are available on request. macOS is in early access and billed per second ($0.000917/s for a full M4 Pro, $0.000467/s for a 7-core VM), on a fresh VM for every job. Dedicated hosts cost more than the shared pool because they're reserved for you around the clock. GitHub prices are GitHub's published rates for comparable machines, checked October 2026. Code review runs only on organizations that switch it on, and pauses (posting nothing) when credit runs out.
What you'd save
Your monthly GitHub Actions minutes, priced at GitHub's list rates and at ours. Every assumption is below and editable — then measure your own jobs before you switch anything.
You'd save per year
—
- GitHub-hosted / month
- —
- compiler.dev / month
- —
- Savings / month
- —
- Machine hours of waiting removed / month
- —
- Engineer time / month
- —
- GitHub list prices, checked October 2026; ours are the published rates above. Discounts or included minutes on your GitHub plan aren't counted.
- Linux speed-up is the median of our benchmarks (8 vCPU vs GitHub 2-core); macOS is not measured yet, so it defaults to 1.0×. Larger Linux sizes are compared like-for-like at 1.0×.
- Your jobs may differ — single-threaded jobs can be slower here. Comparison mode runs your jobs on both and measures them before you switch anything.
- Machine hours removed is GitHub minutes minus our minutes. Engineer time counts only the share you enter.
Try it on your own jobs
A few minutes to set up. Compare on your own jobs, then switch.