Semaphore is an open source CI/CD platform built for engineering teams and coding agents, self-hosted or run in its usage-based cloud with no per-seat fees.
Semaphore runs the build, test and deploy pipelines that start whenever a developer pushes code or opens a pull request. Darko Fabijan and Marko Anastasov began building it in 2011 and released it a year later, and Semaphore Technologies doo still owns and develops it. The current release treats coding agents as first-class users. A developer tells Claude Code or Codex what CI to set up, and the agent writes and runs the configuration. Each pipeline is a YAML file in the repository's .semaphore folder, and the Workflow Builder models it in a graphical editor that generates the YAML. Jobs fan out through job parallelism and a job matrix, share files as artifacts and reuse dependencies through the cache. Promotions chain pipelines together, deployment targets decide who may deploy and keep a history, and Tasks run pipelines on a cron schedule. Test Reports collect results per test and flag flaky tests, while Project Insights and Organization Health chart duration and pass rate. For agents there are the sem-ai tool, the MCP server and a sandbox that checks a change before it is pushed. Semaphore suits engineering teams that want an agent to set up, diagnose and speed up their CI rather than tending YAML by hand. Platform engineers running monorepos use change_in rules so a commit builds only the services it changed. Security reviewers find SAML SSO, role-based access control, audit logs and a SOC 2 Type II certification. Teams choose between Semaphore Cloud on machines the company runs and two self-hosted editions: the open source Community Edition and the licensed Enterprise Edition. A common job is running a test suite in parallel on every pull request from GitHub or Bitbucket, then promoting the green build to staging. Production deploys to AWS or Google Cloud use short-lived OpenID Connect credentials instead of stored keys. Hosted Linux and Apple machines run the cloud builds, and self-hosted agents bring Windows machines or a Kubernetes cluster into the same pipeline. When a build goes red, an agent connected through the MCP server reads the failure and proposes the fix.
Pricing
View plans-
Semaphore Cloud
Free
- Billed monthly by usage per machine minute, no base fee; annual billing at a discount on request
- $15 of credits every month at no cost, about 2,000 Ubuntu x64 2-vCPU minutes
- Ubuntu ARM $0.003 a minute on 2 vCPU, $0.006 on 4 vCPU, $0.012 on 8 vCPU
- Ubuntu x64 $0.0075 a minute on 2 vCPU, $0.015 on 4 vCPU
- macOS $0.09 a minute on 4 vCPU
- Self-hosted agents $0.0025 a minute on any machine
- No per-seat fees
- 20 concurrent jobs by default, higher limits on request
- Artifacts include 20 GB a month of egress and 100 GB a month of storage
- Extra artifact use $0.35 per GB of egress and $0.07 per GB of storage
- 9.6 GB of cache storage per project
- Dedicated Cache Server add-on $400 a month, 10 Gbps bandwidth and 400 GB storage
- Private Docker Registry add-on $400 a month, 10 Gbps bandwidth and 400 GB storage
- No hosted Windows machines; Windows builds run on self-hosted agents
-
Community Edition
Free
- Open source under Apache 2.0, installed on your own infrastructure
- Kubernetes or single-machine install
- Advanced workflows, audit logs and advanced role-based access control are Enterprise Edition features
-
Enterprise Edition
Custom pricing
- License priced on request
- Source-available, self-hosted
- Advanced workflows, audit logs and advanced role-based access control
-
Support plans (add-on)
Custom pricing
- Monthly support add-on in four levels, priced per level below
- Basic $50 a month, email in business hours, no guaranteed response times
- Priority Response $250 a month, email and Slack, best-effort accelerated replies 24/5
- SLA Standard $500 a month, guaranteed responses 24/5 with P1 in 4 hours and P4 in 1 week
- SLA Premium $750 a month, priority queue 24/5 with optional 24/7, P1 in 1 hour and P4 in 3 days
- Business hours are Monday to Friday, 7am to 8pm UTC
-
Success and engineering services (add-on)
Custom pricing
- Monthly success add-on in four levels, plus engineering hours bought in blocks
- Visibility $125 a month, annual optimization session and quarterly group Q&A
- Performance $300 a month, adds private sessions and guided onboarding
- Scale $1,000 a month, adds an annual business review and team training
- Alliance $2,000 a month, adds roadmap discussions and custom terms
- Engineering Access 3 hours $900, 5 hours $1,250, 10 hours $2,000
Features
- Included: Pipeline as code in the repository
- Included: Pipeline triggers on push and pull request
- Included: Parallel jobs and matrix builds
- Included: Dependency caching between pipeline runs
- Included: Self-hosted runners
- Not included: Hosted macOS, Windows and Linux runners
- Included: Scheduled pipelines
- Included: Manual approval gates before deploy
- Included: Protected deployment environments
- Not included: Masked secrets for pipeline jobs
- Included: OIDC cloud credentials for jobs
- Not included: Reusable pipeline templates
- Included: Test reports with flaky test detection
- Included: Artifact passing between jobs
- Included: Pipeline duration and success analytics
- Not included: Merge queue
- Included: Free monthly build minutes
- Included: Self-managed CI server edition
Additional Features
- YAML pipelines in the repository's .semaphore folder
- Workflow Builder graphical pipeline editor
- Job parallelism and job matrix
- Promotions between pipelines
- Deployment targets with Who can deploy rules
- Tasks for cron-scheduled and manual pipelines
- Cache store and cache restore commands
- Workflow and job artifacts
Best for
- Engineering teams that want coding agents to set up and repair their CI
- Teams that would rather pay per machine minute than per seat
- Companies that must run the whole CI platform on their own infrastructure
- Teams that want an open source CI server they can read and audit
- Platform teams running monorepos with many services
- Teams deploying to AWS or Google Cloud without stored access keys
- iOS and macOS teams that need hosted Apple build machines
- Teams with Windows builds willing to run their own Windows agents
- Organizations whose security review asks for SAML SSO, audit logs and SOC 2 Type II
Use cases
- Running tests on every push and pull request
- Splitting a slow test suite across parallel jobs
- Building one job across a matrix of language versions
- Promoting a green build to staging and then production
- Restricting who can deploy to production
- Running nightly pipelines on a cron schedule
- Finding and tracking flaky tests
- Tracking pipeline duration and pass rate over time
- Deploying to AWS with short-lived OIDC credentials
- Self-hosting CI on Kubernetes
- Letting a coding agent diagnose a failed build
- Building iOS apps on hosted macOS machines
Screenshots & Videos
Explore Semaphore in action
User Reviews
Compare with Similar Software
Click to compare side by side with Semaphore
