Semaphore Uncut

Semaphore
  • 5.0 (2)
  • Technology
  • Semimonthly

Semaphore Uncut takes you behind the scenes of building AI-native Semaphore CI/CD. Tune in for CI/CD best practices for AI-driven software development, product updates, and learnings from our engineering team. semaphoreio.substack.com

  1. Nov 6

    Semaphore Open Source 1.6: MCP Server, Partial Job Reruns, and a First Look at a New Workflow Page

    Semaphore Open Source 1.6 is out in both the Community and Enterprise editions, and we have an early look at something new for big pipelines. Here’s what’s new. AI agents can now work with Semaphore Open Source. Every Semaphore Open Source 1.6 install comes with a built-in MCP server, turned on by default. AI agents like Claude Code, Codex, VS Code and GitHub Copilot can read your workflows, pipelines, logs, test results and artifacts, and run or rerun builds. Ask your agent “why did the last build fail?” and it digs through the logs for you. Your CI/CD data stays on your own infrastructure. Rerun only the jobs that failed. When a pipeline fails, only the failed jobs run again. Jobs that passed carry over with their logs, artifacts and test results. You get back to green faster, and you stop spending compute on work that already passed. More of Semaphore is available over the API. New endpoints cover artifacts, test results and organization management, so scripts and agents can fetch build outputs and manage members, roles and service accounts without the UI. On the Enterprise Edition, you can also manage pre-flight checks through the API and use richer OIDC tokens for tighter cloud deployment policies. GitHub merge queues just work. Pull request checks no longer hang, statuses arrive in order, and an auto-cancelled run can no longer overwrite a newer result. A release security teams can approve. Images and the Helm chart are signed with cosign, bundled storage now runs on RustFS with credentials generated per install, and sign-in moved to Keycloak 26.8. Read the upgrade guide before you upgrade. Stored artifacts, job logs and cache entries aren’t migrated to the new storage, and the Keycloak migration is one-way, so back up first. A big thank you to our contributors. 25 people contributed to Semaphore Open Source 1.6, and 13 of them were contributing for the first time. 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

    Semaphore Open Source 1.6: MCP Server, Partial Job Reruns, and a First Look at a New Workflow Page

  2. Oct 27

    Meet SuperPlane 1.0: The Open Source Software Factory for Teams

    For more than ten years, we’ve helped engineering teams ship faster with Semaphore. We watched CI/CD go from optional to obvious. Now we think the same thing is about to happen with coding agents. Every engineer has an agent, but everyone runs it their own way. And someone still has to decide if a task is ready, check the result, chase failing CI and answer review comments. That’s where the hours go. So today we’re launching SuperPlane 1.0: an open source, model-agnostic software factory that gives your whole team one shared, repeatable path from ticket to merged pull request. 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

  3. Oct 9

    Semaphore Open Source 1.6: MCP Server, sem-ai Out of the Box, and Partial Job Reruns

    A new open source release of Semaphore is almost here. Version 1.6 lands exactly one year after 1.5 and packages up around 494 pull requests of work from the past year. Here’s what’s inside. AI and automation, built in The MCP server now ships in the open source editions. It’s included in both the Community Edition and the Enterprise Edition. sem-ai works out of the box. It’s fully supported with no additional configuration needed, and works with both service accounts and user tokens. Almost everything is now configurable through the API. New public API endpoints cover artifacts, plus pre-flight checks in the Enterprise Edition. Between sem-ai and the regular API endpoints, you can now configure almost everything in Semaphore programmatically. Pipeline improvements Partial job reruns are here. For very long and complex pipelines, not everything needs to be rerun anymore. New filtering options in the UI. Finding what you need in your pipelines is now easier. Under the hood RustFS replaces MinIO as the artifact storage engine. With MinIO moving away from the open source community, we switched to RustFS. It also powers our tests. Security fixes and improvements. The release includes a lot of security fixes and general improvements across the board. What’s Coming The release candidate will be ready in the next couple of days, along with the full extended changelog. Get involved We accept PRs from the community. Open an issue or a PR on the Semaphore repository and share your comments and suggestions. All contributions and feedback are very welcome. * Try sem-ai * Try Semaphore Cloud * All product news Till the next time, Pete Miloravac 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

  4. Sep 16

    sem-ai 0.4.0 gives AI agents safer and more efficient access to CI/CD

    AI coding agents can generate and modify code quickly. But to work effectively with CI/CD, they need focused information, clearly defined permissions, and safeguards that prevent unsuitable changes from progressing. The new sem-ai 0.4.0 release improves all three areas. In our latest product news video, Nick demonstrates more compact pipeline responses, flexible context switching, and the ability to create and validate Semaphore preflight checks directly with an AI coding agent. Get the essential pipeline information Requests for Semaphore workflows and pipelines now return a much more compact response by default. In the demonstration, the new response was approximately 99 percent smaller than the previous full output. This gives an AI coding agent the essential pipeline information without filling its context window with unnecessary detail. The complete output is still available when needed. Results can also be limited or filtered by date. This makes sem-ai more efficient by default while giving developers control over how much information the agent receives. Switch between permission contexts sem-ai 0.4.0 also introduces a context flag. You can define separate contexts for different Semaphore organizations or service accounts, then select the appropriate one through a command line flag or an environment variable. For example, one service account could have read only access for reviewing code and investigating pipeline failures. Another could have permission to modify Semaphore configuration. This supports a safer, least privilege approach. An AI agent receives the access required for its current task without automatically receiving broader control. Create preflight checks with your coding agent The release also adds support for Semaphore’s preflight checks API. Preflight checks evaluate a request before a workflow starts. Teams can use them to enforce organizational rules, including required branch naming conventions. In the video, Claude Code creates a preflight check for a Semaphore project. It configures a branch naming rule, adds the check, and tests it against several branches. Workflows from accepted branches continue normally. Branches that do not follow the required pattern are blocked before the pipeline begins. This gives teams another way to turn delivery policies into enforceable safeguards, directly from the development environment where the agent is working. One limitation remains. The init job configuration used by preflight checks cannot yet be modified programmatically. The Semaphore team plans to address this in a future update. sem-ai 0.4.0 is available now Plugins for Claude Code and Codex will update automatically. If you use the sem-ai command line interface directly, it will notify you when the new version is available and provide the command needed to install it. Watch the full demonstration to see sem-ai 0.4.0 in action. Explore the project on GitHub, or create a Semaphore account to try it with your own pipelines. 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

  5. Aug 12

    Rerun Only the Jobs That Failed

    A pipeline fails on a single job, and you shouldn’t have to rebuild everything around it to recover. That’s what changes with this update. Job rerun is here, and it reruns only the jobs that actually failed. Until now, when a job inside a block failed and you reran the pipeline, every job in that block got rebuilt, including the ones that had already passed. The new default rebuilds only the failed jobs. Every other job in the block, plus jobs from earlier in the pipeline, is carried over and reused, so your workflow topology stays intact and only the new jobs get built. The carried-over jobs are excluded from your billable time. That means a rerun costs you only the jobs that failed, not the whole block. Faster turnaround on failures, and a lower bill to match. The old behavior is still one flag away. We exposed a YAML flag called block rebuild that keeps the full block rebuild when you want it. The new, leaner path is the default, and the escape hatch is there when you need it. sem-ai already understands the new behavior, so agents pick it up and rerun the right way automatically. What’s Coming Job rerun is the first building block of doubling down on CI for the second half of 2026. We’re working through what CI looks like in an agent-driven software development world, with sem-ai as the agent-native interface to the CI. Expect more agentic CI capabilities announced through September. * Try sem-ai * Try Semaphore Cloud * All product news Till the next time, Pete Miloravac 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

    Rerun Only the Jobs That Failed

  6. Jun 19

    Flaky Test API Now GA, New Auto-Fix Skill, Skill Quality Improvements

    Flaky test data is now accessible through the sem-ai API, and this week also brings a skill that uses that data to fix the tests automatically. Flaky test data is now generally available through the API. You can query your flakiest tests programmatically, sorted by number of disruptions, with failure timestamps and logs included. Find it on github. A new sem-ai skill can identify and fix flaky tests end-to-end. The agent pulls the highest-disruption tests, gathers failure context, determines the root cause, and writes a fix. It then verifies the result, first by running tests locally, and if that’s not possible, by spinning up Semaphore test boxes to run the test across multiple machines in parallel. Running the test repeatedly across machines is especially important for flaky tests, since a single passing run isn’t enough to confirm a fix. Benchmarking on a real open source project with Claude Opus on high effort showed a cost of $1 to $1.50 per fix. Four existing skills were improved with additional examples. Agents were occasionally not following skill instructions due to a lack of examples. Adding concrete examples improves adherence and makes sem-ai’s output more consistent across runs. What’s Coming User and organization management will be added to the API in an upcoming release. The team is also continuing to refine skills and commands based on usage patterns. * Try sem-ai * Try Semaphore Cloud * All product news Till the next time, Pete Miloravac 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

    Flaky Test API Now GA, New Auto-Fix Skill, Skill Quality Improvements

  7. Jun 4

    Codex Support, Faster Task Creation, and Flaky Test Visibility

    Here’s an update on what our engineering team has been shipping in sem-ai and what’s coming up next. Codex marketplace support is live.A few issues were blocking some users from getting sem-ai working on Codex. Those are now fixed. If you tried it before and hit a wall, it’s worth giving it another shot. Task creation got faster.We cleaned up some unnecessary API calls that were causing hiccups. You can now create and manage tasks on your projects entirely through the CLI, exactly where your agent already lives. sem-ai plugin submitted to the Claude marketplace.We’re in review. We’ll share more once it’s through. On the horizon:The team is working on full platform management through sem-ai. Members, organizations, and everything else currently in the Semaphore UI will be accessible directly through the CLI. Also shipped: Show/hide skipped blocks in the workflow editor is now generally available. If you work with complex pipelines, it makes reading results a lot cleaner. 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

    Codex Support, Faster Task Creation, and Flaky Test Visibility

  8. May 27

    Building an AI-Native CICD Experience with sem-ai

    We’ve been talking a lot recently about AI-native developer workflows and where CICD is heading. This week, we shared one of our most exciting demos yet: a look at how SEMai is changing the way developers interact with CICD systems. In the demo, Marcos starts with a completely clean repository: no CI configuration, no GitHub Actions workflows, nothing. Then, inside Claude Code, he runs a single command: /sem-ai init From there, sem-ai analyzes the repository, detects the tech stack, and generates a tailored CICD pipeline automatically. It recommends: * Linting * Security scanning * Matrix testing * Pipeline topology improvements But the most interesting part comes afterward. Marcos tells the agent: “Work until the pipeline is green.” The system monitors the pipeline, analyzes failures, applies fixes, reruns workflows, and eventually gets the build passing successfully while summarizing everything it changed along the way. Nick also shares lessons learned while building the onboarding experience itself, including why slash commands became critical for reliable agent workflows and how improving contextual skills dramatically increased success rates. This is what we mean when we talk about AI-native CICD:not AI bolted onto existing tooling, but a fundamentally different developer experience built around collaboration between developers and agents. In the full demo, Marko, Marcos, and Nick walk through: * AI-assisted CICD onboarding * Slash command workflows * Self-healing pipelines * Pipeline optimization * The evolution of SEMai skills and agent context We’re excited about where this is heading and look forward to sharing more soon. Pete Miloravachttps://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

    Building an AI-Native CICD Experience with sem-ai
See All (165)

Ratings & Reviews

5
out of 5
2 Ratings

About

Semaphore Uncut takes you behind the scenes of building AI-native Semaphore CI/CD. Tune in for CI/CD best practices for AI-driven software development, product updates, and learnings from our engineering team. semaphoreio.substack.com

Information

  • Creator
    Semaphore
  • Years Active
    2019 - 2026
  • Episodes
    165
  • Rating
    Clean
  • Copyright
    © Semaphore
  • Show Website
    Semaphore Uncut