Notes / Projects / Platform Shipsolid / 08 Strategy Planning

AIOps Overview

The **h-aiops** pillar is an experimental sandbox for the in-house SRE Agent and related AIOps

Updated May 1, 2026 · §202603260022-9 ·

AIOps & Intelligent Operations

The h-aiops pillar is an experimental sandbox for the in-house SRE Agent and related AIOps patterns. It is not part of the production control plane.

The current repo state supports interface design, playbook authoring, MCP configuration, and knowledge-base curation. It does not yet provide a production-ready agent runtime.

What is AIOps?

AIOps applies AI reasoning to operational workflows — root cause analysis, incident triage, infrastructure maintenance — that traditionally require significant human effort and context-switching across multiple tools.

Architecture

┌─────────────────────────────────────────────────────┐
│                    SRE Agent                         │
│                                                      │
│  ┌──────────┐  ┌───────────┐  ┌──────────────────┐  │
│  │ Playbook │  │   Agent   │  │  Knowledge Base  │  │
│  │  Engine  │──│   Core    │──│   (RAG Context)  │  │
│  └──────────┘  └─────┬─────┘  └──────────────────┘  │
│                      │                                │
│              ┌───────┴────────┐                       │
│              │  MCP Transport │                       │
│              └───────┬────────┘                       │
└──────────────────────┼──────────────────────────────┘

        ┌──────────────┼──────────────┐
        │              │              │
  ┌─────┴─────┐ ┌─────┴──────┐ ┌────┴──────┐
  │  Grafana  │ │ ServiceNow │ │   Infra   │
  │MCP Server │ │ MCP Server │ │MCP Server │
  └───────────┘ └────────────┘ └───────────┘

Pillar Contents

DirectoryPurpose
01-sre-agent/Core agent — Python, Claude Agent SDK, MCP client
02-playbooks/Declarative YAML workflows for operational tasks
03-knowledge-base/Runbooks, architecture docs, postmortems for RAG context
drafts/Experimental work

Design Principles

  1. MCP-first — External systems accessed via MCP servers only
  2. Playbook-driven — Workflows declared in YAML, not hardcoded
  3. Human-in-the-loop — Destructive actions require approval
  4. Observable — Agent emits OTel traces to Grafana Cloud
  5. Knowledge-grounded — Curated docs improve agent reasoning

Local graph

Full graph →