scout-mcp

Expose read-only Kubernetes and GitOps state to AI agents via MCP.

14|Updated Jan 17, 2026
One-click install
npx skills add https://github.com/confighub/cub-scout --skill scout-mcp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: scout-mcp
Source: https://github.com/confighub/cub-scout/tree/main/skills/scout-mcp
Command: npx skills add https://github.com/confighub/cub-scout --skill scout-mcp

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

cub-scout turns live Kubernetes and GitOps state into structured, read-only evidence, and this skill makes that evidence accessible to AI agents through MCP or as a deterministic context snapshot.

Core Features & Use Cases

  • MCP tool gateway for read-only verbs: Exposes cub-scout’s observe/diagnose/explain/trace/scan/compare-style capabilities as MCP tools for Claude, Codex, and other MCP-aware agents without mutating cluster state.
  • Deterministic context packs for LLM reasoning: Generates a one-shot JSON snapshot of cluster context and attribution evidence suitable for prompt caching and offline consumption.
  • Standalone vs connected modes: Supports cluster-only snapshots and optionally includes ConfigHub linkage when authenticated, while keeping outputs read-only.

Quick Start

Use the scout-mcp skill to start an MCP server by running: cub-scout mcp serve.

Frequently Asked Questions about scout-mcp

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I expose Kubernetes cluster state to an AI agent via MCP?

To expose Kubernetes cluster state to an AI agent via MCP, launch the read-only MCP gateway using the serve command, which registers cub-scout's observe, diagnose, and trace verbs as tools without mutating cluster state.

Can I capture Kubernetes and GitOps context for offline LLM reasoning?

Yes, you can capture Kubernetes and GitOps context for offline LLM reasoning by generating a deterministic JSON context pack, which provides a stable, one-shot snapshot of cluster evidence suitable for prompt caching and CI environments.

Does the MCP Kubernetes gateway support write operations or mutations?

The MCP Kubernetes gateway does not support write operations or mutations, as it strictly enforces read-only observability by exposing explanation and diagnosis verbs to ensure cluster state remains safely unchanged during AI agent integration.

How do I start an MCP server for read-only Kubernetes observability?

You start an MCP server for read-only Kubernetes observability by running the cub-scout mcp serve command, which initializes the gateway for MCP-aware agents like Claude and Codex to access cluster context safely.

Do I need ConfigHub authentication to use the MCP context packs?

You do not need ConfigHub authentication to generate standalone MCP context packs, but authenticating with cub enables connected-mode linkage to include additional ConfigHub context while maintaining read-only outputs.

What is the best way to debug MCP tool registration for Kubernetes observability?

The best way to debug MCP tool registration for Kubernetes observability is to use the scout-mcp gateway, which exposes deterministic JSON outputs and read-only verbs to verify agent integration and tool availability without cluster risk.