Semaphore MCP for AI. Control every CI/CD step from chat.

Semaphore MCP Server - CI/CD Automation via AI

This server hooks your Semaphore CI/CD account right up to your favorite agent. You ditch switching between web dashboards and terminals; you manage every single part of your delivery lifecycle using natural conversation with your AI client.

Monitoring Builds and Debugging Failures

You've got a broken build, right? No sweat. To check the status of everything, you can run list_pipelines to see all project runs, or use list_workflows for defined workflows. For deep dives, you get summary details on one specific job with get_job, and if that's not enough, you pull every log line using get_job_logs.

You can also check the overall status of a whole pipeline run via get_pipeline, or focus just on a workflow’s context using get_workflow.

If something fails, don't start over. You can use rebuild_pipeline to re-run only the broken sections of an existing job, and you can always stop a runaway process cleanly; that includes stopping a whole pipeline run with stop_pipeline, halting a specific workflow execution using stop_workflow, or just stopping a single running job instantly via stop_job.

Managing Deployments and Targets

You need to release code, period. This server gives you total control over deployment targets. You can list every target associated with the project using list_deployment_targets, and if you know its name, you find it with get_deployment_target_by_name; otherwise, you pull all the specs using get_deployment_target by ID. The lifecycle management is simple: use create_deployment_target to set up a new destination, then activate it immediately with activate_deployment_target, or take it offline entirely with deactivate_deployment_target.

If something changes—say, the URL or credentials—you modify it using update_deployment_target; if you're done with it, delete_deployment_target wipes it out permanently. You can check what happened before by pulling a full history of events with get_deployment_history, and to force an immediate release into a specific environment, you trigger a promotion using trigger_promotion.

If the target is bad news, you delete it; if you just need to know how it looks, you use get_deployment_target.

Artifacts and Code Integrity

Storage management is key. You can see all the built files available for a project scope by running list_artifacts. When you're done with storage, you set the rules defining artifact cleanup using configure_artifact_retention, and you check what those current policies are with get_artifact_retention.

Before committing anything, don't just trust it. You run validate_yaml to make sure your pipeline definition is syntactically correct, preventing messy commits down the line. For quick, isolated operations that don't require a full CI/CD cycle, you execute them right now using trigger_task.

Agent and System Management

The infrastructure runs on agents, so keeping track of 'em is critical. You can see every agent type configured across your organization by checking list_agent_types, or get the full details for one specific definition with get_agent_type. To spin up new resources, you use create_agent_type to define a new self-hosted agent type, and if you need to adjust its settings, you modify it using update_agent_type; if it's obsolete, delete_agent_type removes it.

You can list all active agents via list_agents, or pull detailed information on one specific instance with get_agent. If an agent type needs to be temporarily disabled for maintenance, you use disable_all_agent_types.

Other Tools At Your Fingertips

You also get tools for viewing promotion records with list_promotions, and if the whole project's YAML changes, you can run a general pipeline restart using rebuild_pipeline or trigger an entire fresh execution of a defined workflow with run_workflow; this is different from just running a single task. You'll find everything you need to manage your entire delivery process through one chat interface.