Semaphore Uncut

As AI agents become part of everyday development workflows, authentication is becoming a critical piece of the puzzle.In our latest product update, we’re sharing a behind-the-scenes look at how we implemented OAuth for Semaphore’s MCP server—and what we learned along the way.You’ll get a practical perspective on:* Why OAuth is essential for MCP and AI agents* The challenges of working with evolving specs and inconsistent agent behavior* What actually works in real-world implementations* Key lessons for building secure, reliable integrationsThis isn’t theory—it’s a real engineering deep dive from our team.👉 Read the full blog post👉 Watch the full videoThis work is part of a broader shift at Semaphore, where we’re extending CICD with AI-powered workflows—while keeping developers in control and reducing manual work.If you’re building with MCP or exploring agent-driven workflows, this one’s worth your time.— Pete MiloravacThe Semaphore Teamhttps://semaphore.io This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

Today we’re introducing an important change to how Semaphore pricing works. CICD is no longer something that runs occasionally during development. For many teams, it’s now always-on infrastructure that powers software delivery.To support this shift, we’re introducing a new Semaphore pricing model designed to keep CICD fast, scalable, and affordable as usage grows.What’s changingThe new model separates compute infrastructure from support and success services, giving teams more transparency and flexibility in how they use Semaphore.Key updates include:Lower machine pricingSemaphore’s fastest runners are now available at significantly reduced prices, starting at $0.0075 per minute.Our new f1 machines replace the previous e1 and e2 runners, focusing the platform on high-performance compute optimized for CICD workloads.Usage-based pricing for self-hosted agentsSelf-hosted agents are no longer billed per developer seat.Instead, usage is priced purely based on compute time at $0.0025 per minute.Support and success plans are now separateInfrastructure pricing now reflects compute consumption only.Teams can choose optional Support and Success plans depending on their operational needs and level of optimization guidance.A modular approach to CICDTeams can now combine:* Compute infrastructure* Support plans for operational reliability* Success plans for performance optimization* On-demand engineering assistanceThis modular model makes it easier to scale Semaphore alongside your CICD usage.For existing customersNothing changes immediately.You can continue using your current plan as-is, and when you’re ready, you’ll be able to switch to the new pricing structure directly in the app.Learn moreYou can explore the full pricing details here:👉 Pricing page👉 Try out SemaphoreIf you have any questions, drop us an email at [email protected] Semaphore Team This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

We’re preparing a new update for the Semaphore MCP server that will make it easier for developers to connect AI agents and developer tools.The focus of this update is authentication.Today, connecting an agent to the MCP server typically requires using a long-lived API token. While this works well, it also means developers need to generate credentials, store them in configuration files, and manage them manually.In our next release, coming next week, we’re introducing OAuth authentication support for the MCP server.This will make connecting agents and developer tools significantly simpler.Instead of generating and storing API tokens, developers will be able to authenticate through a familiar OAuth flow. When configuring an agent, a browser window opens, you log in, and approve access to the MCP server. Once approved, the connection is established automatically.This approach removes the need to manage long-lived credentials and makes integrations easier to set up.It also improves compatibility with modern agentic development tools. Some tools have limitations when working with static API tokens, and OAuth removes those barriers.Read more on our blog.Pete MiloravacThe Semaphore Teamhttps://semaphore.io This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

🚀 We’re introducing AI-driven onboarding and workflow automation in Semaphore. Our new assistive agent helps teams generate pipeline configurations, understand failed jobs, re-run workflows, and refine CI/CD setups using natural language — all while keeping developers fully in control. This is the beginning of Agent-Driven Semaphore: CI/CD at the core, AI handling the repetitive work. Watch to see what’s coming next.Give Semaphore a spin:https://semaphore.io/ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

Over the past few weeks, we’ve been refining our roadmap for 2026 and focusing on where Semaphore can deliver the most value.Our direction is clear: extend CI/CD beyond execution — and help developers ship software faster.Here’s what’s coming next. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Semaphore Uncut, Malcolm Matalka (Terrateam) discusses GitOps, Terraform tooling, OpenTofu workflows, and building developer tools that stay invisible. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Semaphore Uncut, Jamie Dobson, co-founder and former CEO of Container Solutions, shares a thoughtful perspective on generative AI, developer productivity, and system stability.The conversation explores why recent research shows engineers feeling happier and more productive with AI tools—while system reliability quietly declines. Jamie explains how AI-generated code can appear correct while introducing subtle long-term risk, and why speed without deep understanding can be dangerous for growing systems.They also discuss why looking back at the history of computing—from time-sharing to early neural networks—helps demystify today’s AI hype. The episode closes with a reflection on the stories we tell about technology, how humans and machines can work together, and what engineering leaders should be paying attention to next.👉 Read the full blog post for the complete conversation — and happy listening! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Semaphore Uncut, Sebastian Gierlinger, VP of Engineering at Storyblok, shares what it really takes to scale a fully remote engineering organization—from early startup days to a 250-person company.The conversation explores how Storyblok scaled engineering across multiple time zones without losing alignment or delivery speed. Sebastian explains why Scrum stopped working as the team grew, what changed after adopting Shape Up, and how specialization—if left unchecked—can quietly turn into a bottleneck for growing teams.They also discuss what remote-first organizations need to get right as they scale, from decision-making and trust to execution at speed. The episode closes with a candid look at how AI is already changing day-to-day engineering work—and why scaling engineering is ultimately less about hiring faster and more about designing better systems.👉 Read the full blog post for the complete conversation — and happy listening! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Semaphore Uncut, Keren Fanan (CEO & Co-founder) and Hadar Geva (CTO & Co-founder) of MyOp share how years of building and scaling frontend systems led them to rethink how UI should be built, shipped, and managed in production.This conversation covers:* How frontend systems break down over time due to framework churn and legacy code* Why treating UI components as runtime contracts changes how teams ship and experiment* How MyOp enables safe UI rollouts and gradual frontend modernization without redeployments* Where this approach goes beyond traditional micro-frontend architectures* How AI-generated UI is accelerating the need for more flexible frontend systems* What MyOp’s open, framework-agnostic model looks like in practiceA clear, practical look at how frontend architecture is evolving—and what it means for teams building long-lived products.👉 Read the full blog post for the complete conversation — and happy listening! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

AI agents can reason, but they don’t actually understand your systems. MCP servers fix that by giving your copilots and assistants structured access to your tools, APIs, and CI/CD data. And the best part is that building one is much simpler than most people expect.In the latest episode of Technical Tips, Tommy walks through how to create a functional MCP server in just a few minutes. He connects it to Semaphore’s API, pulls real project data, and shows how to expose those insights to tools like OpenAI Codex. It’s a straightforward, hands-on demo that finally makes MCP feel practical instead of theoretical.If you’ve been curious about MCP or you’re trying to make your AI tools genuinely useful in day-to-day engineering work, this is the perfect place to start. You’ll see how an MCP server communicates with AI clients, how to wrap a real API into it, how to test everything using the MCP Inspector, and how easily you can turn your CI/CD workflows into a conversational experience.By the end, you’ll understand how surprisingly little code it takes to give your AI agents real operational awareness.Access the full guide on our blog. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Semaphore Uncut, Jonathan Reimer—founder of Crowd.dev and now VP of Outbound Products at the Linux Foundation—breaks down what he’s learned from years inside the open source startup ecosystem.This conversation covers:* How Crowd.dev grew into a core insights platform for OSS communities and was later acquired by the Linux Foundation* The three types of open source companies Jonathan identified after speaking with 400–500 founders* Why GitHub stars are vanity metrics—and which signal actually reflects real adoption* When open source provides a real go-to-market edge, and when it doesn’t* How AI is already influencing OSS development, from automated bug fixes to running LLMs locally* Why open source models may ultimately dominate the model layer as costs fallA sharp, fast overview of the forces shaping developer tools, open source ecosystems, and AI infrastructure today.👉 Read the full blog post for the complete conversation — and happy listening! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Technical Tips, Tommy shows you how to run tests across multiple environments without duplicating your CI/CD setup or writing repetitive jobs.He explains how job matrices in Semaphore help you expand test coverage, catch regressions early, and keep your pipelines organized and efficient.You can also check out the blog post Job Matrix: Feature Showcase for a detailed walkthrough with examples in both the Workflow Editor and YAML.Happy building,The Semaphore Team This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode, Darko welcomes Brian Douglas, Head of Developer Experience at Continue and longtime open source advocate. They talk about the rise of the AI engineer, how AI agents are reshaping developer workflows, and what’s next for open source infrastructure. Enjoy the episode!Read the blog post: https://semaphore.io/blog/brian-douglasLike this episode? Be sure to leave a ⭐️⭐️⭐️⭐️⭐️ review on the podcast player of your choice and share it with your friends.🎙️ This podcast is powered by Semaphore! 🚀Show Semaphore Some Love ❤️⭐ Star us on GitHub: https://github.com/semaphoreio/semaphore🚀 Join us on discord: https://discord.com/invite/FBuUrV24NH⬆️ Upvote on Product Hunt: https://www.producthunt.com/products/semaphore-open-source-ci-cdLet’s make open-source CI/CD even better—together! ✨ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Technical Tips, Tommy walks you through three core principles of automation and CI/CD. These are the foundations that help teams deploy faster, avoid stress, and keep pipelines running smoothly.Learn more: https://semaphore.io/blog/pipelines-explained-principles-ci-cdLike this episode? Be sure to leave a ⭐️⭐️⭐️⭐️⭐️ review on the podcast player of your choice and share it with your friends.🎙️ This podcast is powered by Semaphore! 🚀Show Semaphore Some Love ❤️⭐ Star us on GitHub: https://github.com/semaphoreio/semaphore🚀 Join us on discord: https://discord.com/invite/FBuUrV24NH⬆️ Upvote on Product Hunt: https://www.producthunt.com/products/semaphore-open-source-ci-cdLet’s make open-source CI/CD even better—together! ✨ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode, Darko welcomes Sarah Novotny, a technology leader with 25+ years of experience. From shaping Kubernetes governance at Google to her current work on AI security with the Coalition for Secure AI, Sarah brings deep insight into how open source and governance shape the future of tech.Like this episode? Be sure to leave a ⭐️⭐️⭐️⭐️⭐️ review on the podcast player of your choice and share it with your friends.🎙️ This podcast is powered by Semaphore! 🚀Show Semaphore Some Love ❤️⭐ Star us on GitHub: https://github.com/semaphoreio/semaphore🚀 Join us on discord: https://discord.com/invite/FBuUrV24NH⬆️ Upvote on Product Hunt: https://www.producthunt.com/products/semaphore-open-source-ci-cdLet’s make open-source CI/CD even better—together! ✨ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com

In this episode of Technical Tips, we break down GPT-OSS — OpenAI’s open-source LLMs you can run anywhere. From local setup to cloud hosting, fine-tuning, and performance tips — here’s everything you need to get started with GPT-OSS.Like this episode? Be sure to leave a ⭐️⭐️⭐️⭐️⭐️ review on the podcast player of your choice and share it with your friends.🎙️ This podcast is powered by Semaphore! 🚀Show Semaphore Some Love ❤️⭐ Star us on GitHub: https://github.com/semaphoreio/semaphore🚀 Join us on discord: https://discord.com/invite/FBuUrV24NH⬆️ Upvote on Product Hunt: https://www.producthunt.com/products/semaphore-open-source-ci-cdLet’s make open-source CI/CD even better—together! ✨ This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit semaphoreio.substack.com