We want to share an update on what our product team is currently working on.
Listen on Semaphore Uncut Podcast 🎙️
From CI Jobs to Assistive Agents
Semaphore has always been a reliable place to run builds, tests, and deployments at scale. Now, we’re expanding that foundation.
We’re developing an AI-powered assistant inside Semaphore designed to:
- Help teams set up projects and pipelines faster
- Understand what happened in a failed job
- Re-run workflows on request
- Suggest fixes based on context
- Generate and validate pipeline configuration
This assistant is assistive, not autonomous. It acts only when prompted by developer intent and operates within the boundaries of existing workflows.
Developers remain in control. The agent executes. View a demo below:
A Simpler Onboarding Experience
One of our first focus areas is onboarding.
Instead of starting with documentation and YAML from scratch, developers can describe:
- Where their code lives
- How it should be built
- How it should be tested
- What kind of workflow they want
The assistant translates that intent into a working Semaphore project and pipeline configuration.
From selecting a repository to landing on a generated workflow, the experience is designed to reduce friction while preserving transparency. Developers can review, edit, and refine everything in the workflow editor.
The goal isn’t to hide CI/CD.
It’s to remove unnecessary setup effort.
Understanding and Acting on Pipeline Context
Beyond onboarding, the assistant can:
- Explain why a job failed
- Summarize what happened in a workflow
- Re-run jobs on request
- Validate configuration files
- Suggest improvements
Conversations retain context within a session. That means follow-up instructions like “run it again” or “improve this workflow” are understood in relation to prior actions.
This moves Semaphore from reacting to commits to responding to developer intent — a key step toward becoming a control plane for modern software delivery.
Knowledge That Improves Over Time
We’re also building structured memory into the system.
The assistant can retain contextual knowledge at different levels, such as organization or project, enabling more consistent recommendations over time.
Why this matters:
- Teams don’t start from zero each time
- Patterns can inform smarter suggestions
- Workflow improvements compound over time
This aligns directly with our core belief: high-quality software is the result of consistent, automated hygiene applied over time.
Safe, Scalable Execution
When automation interacts with code, it must be secure and controlled.
Our approach ensures that actions like generating patches, running code, or validating configurations happen within controlled environments tied to Semaphore workflows.
Nothing runs implicitly.
Everything is tied to developer intent.
Transparency and inspectability remain central design principles.
What Comes Next
We’re currently refining:
- The onboarding flow powered by natural language input
- Workflow editing tied to ongoing agent conversations
- Additional automation capabilities around test validation and improvement
Our immediate milestone is delivering a smooth “repository → working workflow” experience. From there, we’ll continue expanding conversational workflow refinement and deeper automation across the software lifecycle.
This is the beginning of Agent-Driven Semaphore.
CI/CD stays at the core.
AI helps automate the work developers don’t want to do.
Developers stay in control.
We’ll share more as we continue building.
Want to discuss this article? Join our Discord.
