Ctrlk

DevOps Agent - Argonaut

Argonaut is an LLM-powered CLI and conversational AI tool that automatically diagnoses, recommends, and helps fix issues with ArgoCD applications — directly within a Slack channel or command line interface or Google chat.

It acts as an intelligent assistant that listens to alerts and user queries, recommends the exact commands needed to debug or resolve an ArgoCD issue, executes them (automatically or on demand), analyzes the output, and continues the diagnostic loop — all within a single Slack thread/ Google chat thread, maintaining full context throughout the conversation. Argonaut bridges the gap between ArgoCD alerts firing and engineers knowing exactly what to do next — replacing guesswork with AI-guided, executable recommendations.

Why Use Argonaut in OpsMx?

ArgoCD is the backbone of GitOps-based continuous delivery for Kubernetes applications. However, when ArgoCD applications fail — sync errors, OutOfSync states, health degradations, missing resources — engineers must manually investigate across multiple tools (kubectl, argocd CLI, git, Prometheus, Elasticsearch) to diagnose and fix the issue.

Argonaut solves this by:

  • Eliminating manual triage — no more guessing which ArgoCD or kubectl command to run

  • Closing the loop between alerts and action — Prometheus, ArgoCD, and Elasticsearch can post directly into Slack; Argonaut picks up those alerts and immediately recommends corrective commands

  • Keeping context across the conversation — the entire diagnostic session lives in a Slack thread, so every recommendation is informed by what came before

  • Enabling LLM-guided GitOps fixes — Argonaut can suggest changes to files in Git, create a PR, and merge it via shortcuts — all from Slack

  • Supporting both manual and automated remediation — teams can run with human approval or enable AUTO-RUN once confident

Scope for the Users

1. Instant, Actionable Diagnostics

Engineers no longer need to know every ArgoCD or kubectl command by heart. Argonaut recommends the exact command — including complex compound commands using && and | — to diagnose and fix the issue immediately.

2. Reduced Mean Time to Resolution (MTTR)

The continuous loop of recommend → execute → analyze → recommend shortens the time from "alert fired" to "issue resolved," reducing the manual investigation cycle that typically spans multiple tools and terminal sessions.

3. Context-Aware Assistance

Because the entire session lives in a Slack thread, every subsequent recommendation is informed by the full conversation history — Argonaut doesn't ask you to repeat yourself or lose track of what has already been tried.

4. Works Where Engineers Already Are

No new portal to log into. Argonaut operates directly in Slack — where alerts already arrive and where engineering teams already collaborate — making adoption frictionless.

5. GitOps Fix Automation

With GIT-FIX, GIT-PR, and GIT-MERGE, engineers can make configuration corrections, raise a PR, and merge it — all from within the same Slack thread — without switching to a code editor or GitHub UI.

6. Flexible Automation Level

Teams can start with full human approval (manual RUN) and graduate to AUTO-RUN as confidence in Argonaut's recommendations grows — giving organizations control over how much autonomy they grant the tool.

7. Full Alert-to-Fix Traceability

With ArgoCD, Prometheus, and Elasticsearch all posting into Slack, Argonaut creates a single, traceable thread from alert fired → diagnosis → commands run → fix applied — a built-in audit trail for every incident.

How It Works — The Argonaut Flow

Key conversation rules:

  • All Argonaut responses begin with NAUT — to clearly identify AI responses

  • All tool execution outputs begin with TOOL

  • Argonaut never responds to messages starting with NAUT — preventing infinite response loops

  • Users can bypass Argonaut and talk to each other by prefixing messages with NAUT

  • Users can run their own commands via: RUN argocd <your command>

Integrations

Built-in CLI Tools

Argonaut has native access to the following executables out of the box:

Tool
Purpose

argocd

ArgoCD application management and diagnostics

kubectl

Kubernetes resource inspection and management

git

Repository operations

gh

GitHub CLI for PR creation and management

yq

YAML processing and editing

jq

JSON parsing and filtering

bash

Standard shell commands and compound operations

User Interface

  • Slack — primary interface for all interactions, alert ingestion, command recommendations, and output display

  • Email — notification delivery

LLM Interface

  • OpenAI GPT-4o and GPT-4.1 via OpenAI APIs — powers all command recommendation and output analysis

Conversation Database

  • File-based — lightweight, local conversation history

  • Elasticsearch — scalable conversation storage for larger deployments

Alert Sources (Slack-integrated)

  • ArgoCD — application sync and health alerts

  • Prometheus — infrastructure and application metric alerts

  • Elasticsearch — log-based alerts

Key Features & Commands

Feature / Keyword
Description

AUTO-RUN

Automatically executes Argonaut's recommended commands without requiring the user to type RUN — ideal once the team has gained confidence in Argonaut's recommendations

RUN

Manual trigger to execute the command recommended by Argonaut in the Slack thread

GIT-FIX

Shortcut to make changes to files in a Git repository

GIT-PR

Shortcut to create a pull request from the current changes

GIT-MERGE

Shortcut to merge an open pull request

SUMMARIZE

Summarizes the entire diagnostic conversation in the current Slack thread

HELP

Lists all available keywords, shortcuts, and concepts

Limitations

  • Configuration and performance of ArgoCD itself — server setup, ArgoCD controller tuning, infrastructure-level concerns cannot be solved by Argonaut.

PreviousScouteSuiteNextRemediation Learning Loop

Last updated