1.9×
Faster Docker builds than GitHub’s runner. A production Docker image in 7m34s on Sengi. 14m36s on GitHub. Same 4 vCPU.
5 GHZ BARE METAL FOR GITHUB ACTIONS.
Change one line: runs-on.
Or let your coding agent do it, with one prompt.
Up to 1.9× faster builds. Up to 3× the multi-core CPU. Every job runs on an AMD Ryzen 9 7950X3D desktop CPU boosting past 5 GHz, on bare metal with local NVMe: up to 3× the multi-core and 1.65× the single-core CPU of GitHub’s runner, and 15× its random-read disk speed. Docker builds finish 1.9× faster. A fresh VM for every job, with 2 vCPU jobs ready in 2 seconds on a pre-booted VM. And it costs less: $0.003 a minute against GitHub’s $0.006, up to 74% less per build. Built for long test suites, Docker builds and the PRs your coding agents open all day.
test:- runs-on: ubuntu-latest+ runs-on: sengi-standard-4-ubuntu-2404 steps: - uses: actions/checkout@v4
AMD Ryzen 9 7950X3D desktop CPUs boosting past 5 GHz, on bare metal, with local NVMe.
A fresh VM for every job; 2 vCPU jobs ready in 2 seconds.
Faster Docker builds than GitHub’s runner. A production Docker image in 7m34s on Sengi. 14m36s on GitHub. Same 4 vCPU.
The multi-core CPU of GitHub’s runner. Every 2 vCPU runner runs on a 16-core AMD Ryzen 9 7950X3D, not a cloud server CPU.
The random-read disk speed of GitHub’s runner. 296k random-read IOPS against 19.6k, and 3× the random writes. Local NVMe on bare metal, not network storage like AWS EBS.
Faster single core. 5 GHz desktop CPUs against the server CPUs behind GitHub’s runners.
Docker build: $0.045 a build on Sengi. $0.175 on GitHub.
All three builds, at 72% less. $0.069 on Sengi for the lot. $0.249 on GitHub.
Faster Docker builds with layer caching. 68 s down to 36 s with a warm type=gha layer cache. Included.
One 5 GHz CPU model on every job we ran. GitHub’s jobs landed on four different server CPUs. Sengi: a Ryzen 9 7950X3D, every run.
Ready in 2 seconds. Your 2 vCPU job lands on a fresh VM that’s already running. Boot time is never billed.
| Build | GitHub | Sengi | Faster |
|---|---|---|---|
| Docker image (PostHog) | 14m36s$0.175 a build | 7m34s$0.045 a build | 1.9× |
| Rust, TypeScript and Docker together | 20m43s$0.249 | 11m28s$0.069 | 1.8× |
| Rust build | 3m53s$0.047 a build | 2m25s$0.015 a build | 1.6× |
| TypeScript typecheck (opencode) | 2m14s$0.027 a build | 1m29s$0.009 a build | 1.5× |
| Job | GitHub | Sengi | Faster |
|---|---|---|---|
| Node + Docker build | 539 s$0.054 a run | 299 s$0.015 a run | 1.8× |
| Vite build and test | 552 s$0.060 a run | 328 s$0.018 a run | 1.68× |
| Mastodon Docker image build | 252 s$0.030 a run | 172 s$0.009 a run | 1.47× |
| fd (Rust) build and test | 76 s$0.012 a run | 51 s$0.003 a run | 1.49× |
| GitHub CLI (Go) test and build | 367 s$0.042 a run | 284 s$0.015 a run | 1.29× |
4 vCPU: RunsOn’s open benchmark harness, run on Sengi October 9, 2026, against RunsOn’s published GitHub results. Measured October 2026 against GitHub’s ubuntu-latest, same 2 vCPU. Medians at list prices.
Up to half the price per minute.
Builds finish up to 1.9× faster.
Together: up to 74% less per build.
| Runner size | GitHub per minute | Sengi per minute | Per minute | Builds finish | Total saved per build |
|---|---|---|---|---|---|
| 2 vCPU / 8 GB | $0.006 | $0.003 | 50% less | 1.9× faster | 74% |
| 4 vCPU / 16 GB | $0.012 | $0.006 | 50% less | 1.9× faster | 74% |
| 8 vCPU / 32 GB | $0.022 | $0.012 | 45% less | 1.5× faster | 64% |
Fewer minutes. Each one cheaper.Builds finish up to 1.9× faster, and every minute GitHub bills at $0.006, Sengi bills at $0.003. Boot and teardown are never billed.
GitHub’s published Linux runner rates, September 2026. Total saved per build: each side’s run time for the same build at its per-minute price. Rust build: 6m50s on GitHub, 3m33s on Sengi. Docker build: 14m36s and 7m34s at 4 vCPU, 9m46s and 6m26s at 8 vCPU.
Two ways to make it. Both end in a pull request you merge. Nothing to run on your side, no tokens to copy.
Agent path
Copy the prompt into your coding agent. No config, no manual edits.
# Migrate GitHub Actions to Sengi runners
Migrate this repository's workflows from GitHub-hosted runners to Sengi runners.
Keep GitHub Actions; change only the `runs-on` line of each supported job, and open
a pull request the user reviews.
**Prerequisites:** `gh` CLI authenticated, and the repository connected to a Sengi
project at https://app.sengi.run — installing the GitHub App alone is not enough,
see Step 1.
## Configuration
| Setting | Value |
| --- | --- |
…
See the whole prompt or curl -s https://sengi.run/agent.md
Manual path
One click installs the GitHub App and connects the repo. No config files, no tokens.
Connect GitHubSwap ubuntu-latest for sengi-standard-2-ubuntu-2404, or the 4 or 8 vCPU label, and merge.
build:- runs-on: ubuntu-latest+ runs-on: sengi-standard-2-ubuntu-2404
Connect the repo, change the label, merge. That’s the whole switch.
No lock-in. Your workflows stay pure GitHub Actions, and switching back is one revert.
Three sizes. Up to 74% less per build than GitHub.
Per minute. No subscription. No minimum.
PREMIUM 2 · MOST BUILDS AND CHECKS
$0.003/ minute
Lint, tests, the checks your agent triggers on every PR.
PREMIUM 4 · BIGGER TEST SUITES
$0.006/ minute
Parallel test suites and bigger builds.
PREMIUM 8 · HEAVY BUILDS
$0.012/ minute
Docker-heavy builds and your biggest jobs.
$0.003, $0.006 and $0.012 a minute for 2, 4 and 8 vCPU, against GitHub’s $0.006, $0.012 and $0.022: up to half the price per minute. Builds finish up to 1.9× faster too, so together you save up to 74% a build. Cache included: your cache steps work unchanged, and Docker builds keep their layers between runs with type=gha. Boot and teardown are never billed. 3,000 minutes included every month; add a payment method to run past them.
YOUR CI BILL
Enter last month’s GitHub job-minutes.
See what the same work costs on Sengi.
Sengi is a managed GitHub Actions runner service: the fastest hardware for your jobs, 5 GHz AMD Ryzen 9 7950X3D bare metal with local NVMe and a fresh VM for every job. You keep GitHub Actions, your workflows and your logs; Sengi runs the jobs up to 1.9× faster, for up to 74% less per build.
Sengi charges $0.003/min for 2 vCPU and 8 GB RAM, $0.006/min for 4 vCPU and 16 GB RAM, or $0.012/min for 8 vCPU and 32 GB RAM. Billed per minute of each job, like GitHub; boot and teardown are never billed. No subscription, no minimum. See pricing and the cost calculator.
No. You change one line per job, the runner name, by hand or with your coding agent. Steps, triggers and secrets stay exactly as they are.
Yes. Once the runner change is merged, every push and pull request runs on Sengi for up to 74% less. The more PRs your agents open, the more you save.
Yes. Change runs-on back, or revert the pull request, and your next job runs on GitHub’s runners again.
Whenever your team pays GitHub for Actions minutes: private-repository CI past your plan’s included minutes, every 4 or 8 vCPU job (GitHub bills those from the first minute), and the steady stream of pull requests your coding agents open. On Sengi those minutes cost half as much, $0.003 against GitHub’s $0.006, and builds finish up to 1.9× faster: together, up to 74% less per build.
Every job on ubuntu-latest or ubuntu-24.04. Sengi is built for Linux x64 on Ubuntu 24.04. The agent prompt finds every one of those jobs for you.
Sengi keeps 2 vCPU runners pre-booted, so your job lands on a VM that’s already running: ready in 2 seconds. Every job still gets its own fresh VM, boot time is never billed, and your dashboard shows each job’s start and run time.
Yes. On RunsOn’s open benchmark harness, run October 9, 2026 at 4 vCPU, Sengi builds a production Docker image 1.9× faster than GitHub’s runner at 74% less per build, and runs the Rust, TypeScript and Docker builds together 1.8× faster at 72% less. Random disk reads are 15× faster and random writes 3×. A job that finishes sooner bills fewer minutes, so the speed cuts the bill too.
Bare-metal machines built like gaming PCs: AMD Ryzen 9 7950X3D desktop CPUs boosting past 5 GHz, DDR5 memory and local NVMe storage, every one picked for single-core speed. Dedicated bare metal with local NVMe, not network-attached storage like AWS EBS. A 2 vCPU runner gets up to 3× the multi-core and 1.65× the single-core CPU of GitHub’s 2 vCPU runner, and a 4 vCPU runner gets 15× the random-read disk speed and 3× the random writes of GitHub’s. Each job gets its own fresh VM.
Yes. Matrix and parallel jobs each get their own fresh VM.
They keep working, with no workflow changes, and they’re included. actions/cache and the setup actions’ caches are served from Sengi’s own cache. Docker builds keep their layers between runs with cache-from: type=gha and cache-to: type=gha,mode=max, and run 1.9× faster warm. See Docker layer caching on Sengi.