# Argo Workflows MCP for AI Agents MCP > Argo Workflows lets your AI agent take full control of complex infrastructure orchestrations running on Kubernetes. You can query, list, and inspect every active pod, workflow template, or scheduled job without touching a command line. It's instant visibility into your entire deployment pipeline. ## Overview - **Category:** loved-by-devs - **Price:** Free - **Tags:** kubernetes, workflow-automation, container-orchestration, ci-cd, infrastructure-as-code, job-scheduling ## Description Running large-scale container deployments means managing incredibly complicated dependency graphs. This MCP connects your AI agent directly to your Argo Workflows cluster, giving you natural language access to every running process and historical record. Instead of opening multiple dashboards or wrestling with `kubectl` commands, you simply ask your agent what’s going on. Your agent can list all active jobs across different namespaces, dive deep into a failed pipeline's resource tree, or check if that critical nightly cron job actually ran this morning. When something breaks—and it always does—you don't waste time figuring out *where* to look; you just ask your AI client. The entire Vinkius catalog makes this possible, giving your agent the deep visibility needed to debug complex infrastructure patterns immediately. ## Tools ### list_workflows Retrieves a list of all active workflow executions running within the Kubernetes namespace. ### get_workflow Fetches the detailed resource tree and current status for any specific Argo workflow instance. ### list_workflow_templates Provides an inventory of reusable WorkflowTemplates that can be used to define new jobs. ### list_cron_workflows Lists all scheduled CronWorkflows, helping you see what recurring jobs are set up in the cluster. ### list_archived_workflows Searches and lists historical workflow records stored in the Argo history database for auditing purposes. ### get_server_info Retrieves basic operational information about the underlying Argo Workflows server itself. ## Prompt Examples **Prompt:** ``` I need to know if the nightly billing process ran successfully yesterday. ``` **Response:** ``` **Workflow Status Report: Billing Process** * **Job Name:** `billing-nightly` * **Status:** ✅ SUCCESS * **Completion Time:** 2026-06-11T04:00:15Z * **Duration:** 18 minutes The run completed successfully. All associated pods (pg_extract, s3_upload) reported a clean exit code 0. ``` **Prompt:** ``` What's wrong with the 'user-sync' workflow? It's stuck in pending. ``` **Response:** ``` I checked the tree for `user-sync`. The failure is at node **Authentication Step**. * **Error:** Timeout waiting for external service credential. * **Code:** 401 Unauthorized (S3) * **Action Needed:** Review the Service Account's IAM policy to ensure it has read access to the target bucket. ``` **Prompt:** ``` List all reusable templates I can use for data processing. ``` **Response:** ``` **Available Workflow Templates in Default Namespace:** 1. `data-cleanup`: Requires S3 Bucket URI input. 2. `model-training-tmpl`: Accepts array of hyperparameters (e.g., learning_rate). 3. `slack-notification-hook`: Takes a target channel and message body. ``` ## Capabilities ### List all current workflows Retrieve a list of workflow executions that are running, pending, or have recently finished within any specified Kubernetes namespace. ### Inspect specific workflow status Get the detailed resource tree and operational status for a single Argo workflow instance to pinpoint exactly where a failure occurred. ### View reusable templates and scheduled jobs List parameterized WorkflowTemplates, allowing you to see which job structures are available, and list all recurring CronWorkflows that handle scheduled tasks. ### Audit historical runs Search through archived workflow records in the history database to understand past infrastructure patterns or identify old failure points. ## Use Cases ### The Nightly ETL Pipeline Failed A data engineer asks their agent to check the 'data-pipeline' workflow. The agent uses get_workflow, identifies that a specific step failed with an S3 permission error, and tells them exactly which parameter needs fixing. ### Auditing Compliance for Scheduled Jobs An SRE needs to know if the billing report ran last month. They ask the agent to list_archived_workflows, quickly finding the historical record and confirming successful execution dates for compliance purposes. ### Checking Available Job Blueprints A developer asks what reusable templates are available. The agent uses list_workflow_templates, providing a clear list of parameterized options like 'model-training-tmpl' so the developer can start coding immediately. ### Debugging a Pending Service Deployment A DevOps engineer sees an active workflow stuck in 'pending'. They ask the agent to get_workflow, and it reports that the issue is with node resource allocation, giving them the exact error message needed for platform adjustments. ## Benefits - Eliminate dashboard hopping. Instead of opening multiple UIs to check status, your agent gives you a single, conversational answer about the health of any workflow. - Pinpoint failures faster using get_workflow. You get the precise resource tree that shows exactly which node or pod failed and why, saving hours of debugging time. - Manage recurring jobs easily. List cron_workflows to audit all scheduled tasks without manually checking the deployment manifest files for every service. - Understand your infrastructure history. list_archived_workflows lets you look back at past runs to understand patterns that might cause future failures. - Save keystrokes and time. Your AI agent handles complex API calls, turning difficult CLI commands into simple chat prompts. ## How It Works The bottom line is that you talk to your AI client like talking to a teammate; it handles all the complex Kubernetes API calls for you. 1. Subscribe to this MCP and provide your Argo Cluster Server URL along with a valid RBAC Bearer Token. 2. Connect your preferred AI client (like Claude, Cursor, or Windsurf) using the Vinkius platform. 3. Ask your agent a question—for example, 'What's wrong with the nightly ETL job?'—and get immediate status reports and failure details. ## Frequently Asked Questions **How can the Argo Workflows MCP help me debug a broken deployment?** It lets your agent check deep into the workflow's resource tree, showing you exactly which step failed and why. You get granular details about the failure—like an incorrect code or missing permission—without having to navigate complex UI screens. **Does this MCP track historical data for Argo Workflows?** Yes, it can list archived workflows. This is huge for compliance and auditing because you don't have to guess what happened last month; the agent finds that record for you. **Is this better than using basic kubectl commands?** Absolutely. While `kubectl` gives status, this MCP provides context. It connects all the pieces—the template definition, the scheduled run, and the live resource state—in one conversational answer. **What if I need to see a list of all recurring jobs?** You can use the MCP to list cron workflows. This gives you a clean inventory of every job that runs automatically, making it easy to audit schedules and check for orphaned or outdated tasks. **Can I see what reusable templates are available in my cluster?** Yes. The agent lists all the workflow templates defined in your namespace. This helps developers quickly find existing blueprints, saving them time writing boilerplate code.