No AI training
Customer content is never sold, used for advertising or competitive analysis, or repurposed for generalized model development.
GitHub Actions-compatible. Independently run.
GitHub Actions can go down. Your delivery pipeline shouldn’t.
Runnable executes your existing workflows with an independent scheduler and clean Linux machines—so GitHub stays your source of truth without becoming a single point of failure for CI.
Scan first, migrate with a reviewable pull request, and run in parallel before switching. Read the migration guide.
10 developers · 3,000 CI minutes
See what $20/month buysRunnable independent CI
$20/month · $0/developer
GitHub Team seats
$40/month · $4/developer
.runnable/workflows/ci.yml
name: CIon: push: branches: [main] pull_request: jobs: test: runs-on: ubuntu-latest strategy: matrix: node: [22, 24] steps: - uses: actions/checkout@v4 - uses: actions/setup-node@v4 with: node-version: ${{ matrix.node }} - run: npm ci && npm testlint4stest (22)18stest (24)17spackage9s
Privacy-first CI
Your code runs here. It never trains here.
Runnable processes customer content only to provide, secure, and support your CI. We never use source code, diffs, workflows, logs, artifacts, secrets, or support attachments to train or improve AI models—ours or anyone else's.
No opt-out required. This applies to every plan.
A straightforward business model. You pay Runnable to run CI. Your development activity is not an advertising asset or a training corpus.
Minimal source custody
Your repository stays in GitHub or Entire. Each job fetches the requested commit onto one clean, single-use runner.
Least-privilege access
Short-lived, read-only repository credentials are refreshed during execution. Platform-wide credentials never enter the job.
Tenant-isolated secrets
Per-organization workspace keys, a fresh data key per secret, authenticated context, and no customer secrets for fork pull requests.
Explicit retention
Logs and artifacts follow published retention windows, so what Runnable keeps—and for how long—is never ambiguous.
Provider-neutral CI
Use GitHub or Entire with familiar workflow syntax. Runnable executes and records CI independently of GitHub Actions.
Everything after git push
The Actions product, rebuilt around your team.
Your Git provider remains the source of truth. Runnable owns the workflow lifecycle from authenticated event to final job conclusion.
GitHub or Entire source
Use the GitHub App, an Entire regional mirror, or register an Entire-native repository with authenticated event delivery.
Independent scheduler
Runnable parses, plans, queues, and coordinates every job without delegating control to another CI service.
Live, searchable logs
Stream masked output in real time, read historical ranges, or let MCP agents diagnose failures and watch deployments on any plan.
Job-scoped security
Ephemeral credentials, envelope-encrypted secrets, fork isolation, and one clean machine per job.
Artifacts and cache
Keep official artifact and cache action compatibility with retention and quotas controlled by your plan.
Usage you can explain
Authoritative machine time, normalized Linux minutes, immutable metering, and reconciliation-ready billing.
Built-in adaptive PR review
PreviewOne review. The depth matches the change.
Configure runnable/pr-review@v1 once. A fixed risk policy routes the same pinned PR revision from a focused first pass to an adversarial conclave, then adds compiler mapping and differential tests when the change needs evidence.
minimum-depth sets the floor—not a separate product. Every path produces one advisory GitHub review, one authenticated report, and stays inside the same hard cost ceiling.
One actionOne pinned revisionHard cost ceiling
.runnable/workflows/pr-review.ymlPreview
name: PR reviewon: pull_request: types: [opened, synchronize, reopened, ready_for_review] permissions: {} jobs: review: runs-on: ubuntu-latest steps: - uses: runnable/pr-review@v1 with: model: recommended max-cost-usd: "5.00" minimum-depth: standardIllustrative resulthigh risk → evidence
Reviewers
4 independent
Finding
1 verified
Coverage
Complete
Tenant scope can be skipped. The generated test passed on the merge-base and failed on the PR head. Security owner review recommended.
Standardlow risk · 1 reviewer
Focused review
One bounded reviewer inspects the changed code and its immediate context without checking out the repository.
Thoroughmedium risk · 2 reviewers
Adversarial round
Independent passes feed a shared candidate pool. Challenges look for counterexamples before a fresh adjudicator decides what survives.
Evidencehigh risk · 4 reviewers
Review plus verification
Roslyn and the TypeScript compiler map the blast radius. Strong hypotheses become secretless tests against the merge-base and PR head.
Explicit adjudication
Every incoming finding gets a disposition.
At evidence depth, each finding from the earlier review is linked back to its original claim or explicitly rejected. Rewording does not duplicate it; rejection does not quietly restore it.
F-01
Confirmed
Survived challenge and remains one finding.
F-02
Rejected
Equivalent behavior or insufficient evidence.
F-03
Verified
Base-pass and head-fail is labeled a verified regression.
Visible risk reasons, advisory output, one $5 default AI ceiling, and $0.25 only after the single review is delivered. Runner compute and taxes are separate.
Browser QA
Gated previewYour build passed. Can your users finish the journey?
Run plain-language cases against your deployed preview in real Chrome. runnable/qa@v1 clicks through the experience and checks the results you expect, right inside your workflow.
Deploy first, pass the preview URL, and keep your QA cases beside the code. Organization owner enablement and separate AI pricing consent are required.
Read the Browser QA post01 · Describe
Plain-language steps. Observable expectations.
Check navigation, sign-in, and other important journeys. Each case starts with a fresh browser session.
02 · Inspect
Passed, failed, or inconclusive.
Per-case summaries and a JSON report show the outcome. Safe final screenshots are included for suites without secrets. QA evidence expires after one day.
03 · Control
Choose the gate. Set the budget.
Fail on failed or inconclusive cases, or collect advisory results. Set an AI cost ceiling for each run; normal runner compute is billed separately.
Built for Entire's Git network
CI/CD for Entire repositories, available now.
01
Connect your repository
Use either a /gh/ mirror or an Entire-native /et/ repository URL.
02
Deliver events
GitHub sends signed events; Entire-native repositories receive a one-time authenticated webhook credential for a relay.
03
Run through Entire
Ephemeral jobs authenticate with ENTIRE_TOKEN and fetch directly from the selected regional cell.
Drop-in where it matters
Move the file. Keep the semantics.
The migration assistant scans your existing workflows, explains unsupported constructs, then opens a pull request that copies compatible YAML for a parallel pilot without rewriting it.
push, pull request, release, issue, schedule, manual, and repository dispatch
matrices, expressions, needs, outputs, conditions, and concurrency
reusable workflows, environments, and approval gates
shell, JavaScript, composite, Docker, and local actions
artifacts, cache, secrets, cancellation, and reruns
An explicit V1 boundary
No silent surprises.
Jobs and steps run for up to six hours. Native services, Compose, Testcontainers, OIDC, and write-scoped GitHub tokens have explicit runtime contracts; job containers and non-Linux architectures remain validation boundaries.
services.full
Native service containers, Docker Compose, docker run, and Testcontainers
share the isolated job Docker daemon.
Independent CI, priced by workload
Your CI bill shouldn't grow with your team.
Runnable is independent delivery infrastructure: its own scheduler, clean Linux machines, masked logs, and hard spend controls. Every developer gets access. Only workload, concurrency, and retention change the bill.
Billing-model lens · headcount economics
10 developers. One Runnable CI bill.
Runnable independent CI
$20
$0/developer · 3,000 minutes
GitHub Team seats
$40
10 × $4/user · 3,000 minutes included
The Runnable total is Hobby plus 2,500 overage minutes, with 1 concurrent job and 7-day logs. Add another developer at the same workload: Runnable adds $0/month; GitHub Team adds $4/month. Move up a Runnable plan for throughput and retention—not headcount.
Migration lens · cost of independence
What $20/month buys when Team stays.
GitHub Team seats
$40
unchanged for code hosting and collaboration
Runnable independent CI
+$20
a separate delivery path for every developer
At 3,000 minutes, GitHub Team's included Actions usage costs $0 incrementally. Runnable's $20/month is an investment in a separate scheduler, clean-machine runner fleet, masked logs, and spend controls—not more GitHub-hosted minutes.
GitHub Team and Runnable are not interchangeable products: Team bundles code hosting and collaboration. GitHub Free organizations support unlimited collaborators and include 2,000 private-repository Actions minutes; at 3,000 Linux 2-core minutes, the incremental Actions usage is $6. Standard runners remain free for public repositories. Both products charge $0.006 per additional comparable minute. Read the full comparison and sources.
Plans by workload
Choose capacity, not seats.
Unlimited users and developers on every plan. Included compute, explicit overage pricing, and customer-controlled spend limits keep headcount out of the invoice.
Hobby
For founders and new startups shipping their first workflows.
$5 / month
then $0.006 per normalized minute
Unlimited users and developers
500 normalized minutes
1 concurrent jobs
7-day log retention
MCP agent access
Starter
For small teams moving their core workflows.
$29 / month
then $0.006 per normalized minute
Unlimited users and developers
3,000 normalized minutes
3 concurrent jobs
14-day log retention
MCP agent access
Pro
PopularFor growing teams running CI in parallel.
$99 / month
then $0.006 per normalized minute
Unlimited users and developers
10,000 normalized minutes
10 concurrent jobs
30-day log retention
MCP agent access
Scale
For high-throughput engineering organizations.
$299 / month
then $0.006 per normalized minute
Unlimited users and developers
30,000 normalized minutes
30 concurrent jobs
90-day log retention
MCP agent access
Enterprise
CustomUnlimited users and developers, SAML SSO, 100,000+ normalized minutes, 100+ concurrent jobs, custom capacity, retention, support, and commercial terms.
Your workflows are already runnable.
Connect GitHub or an Entire-native repository and see the compatibility report.