harvester-mcp-server
Provides a Go-based Model Context Protocol (MCP) server that lets AI assistants interact with a Harvester HCI cluster by translating MCP requests into Kubernetes/Harvester API operations (CRUD-like actions for selected core and Harvester-specific resources) and returning human-readable, LLM-friendly formatted results.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Security is primarily determined by kubeconfig permissions to the target cluster. The MCP server’s external transport/security controls (TLS, authZ/authN for MCP requests) are not described, so assume it is intended for trusted environments. kubeconfig handling is via standard kubeconfig paths/flags; no explicit guidance is provided on preventing logging of sensitive info or enforcing least privilege, so risk depends on runtime configuration and logging.
⚡ Reliability
Best When
You run a single-tenant MCP server locally (or in a trusted environment) with a kubeconfig that has the minimum required permissions, and you primarily need read-only (list/get) cluster inspection plus occasional deletes.
Avoid When
You need fine-grained authorization/auditing at the MCP tool level, you require strong guarantees around destructive operations, or you plan to expose this server to untrusted networks/users.
Use Cases
- • Use Claude Desktop/Cursor to list and inspect Kubernetes/Harvester resources (pods, deployments, services, namespaces, nodes, CRDs).
- • Query Harvester-specific resources such as virtual machines, images, volumes, and networks.
- • Assist operators with natural-language investigation of cluster state and summaries grouped by namespace/status.
- • Build conversational workflows for cluster read operations (and limited delete operations as documented).
Not For
- • Performing fully automated infrastructure changes safely without operator oversight (no strong safety/permission controls are described beyond kubeconfig auth).
- • High-throughput or public multi-tenant access (this is a local/server process that uses a cluster kubeconfig).
- • Services requiring a documented web REST/GraphQL API, SDKs, webhooks, or rate-limit guarantees typical of SaaS APIs.
Interface
Authentication
Authentication is delegated to Kubernetes via the provided kubeconfig. No additional auth layer for the MCP server is documented (e.g., no API key, no TLS termination, no MCP auth).
Pricing
Open-source tool (license Apache-2.0) with self-hosted infrastructure costs only; no pricing model described.
Agent Metadata
Known Gotchas
- ⚠ This is backed by Kubernetes API semantics; agent prompts that request unsupported verbs/resources may fail depending on implemented tools.
- ⚠ Destructive operations (delete) are documented for some resource types; an agent may attempt deletes if not constrained.
- ⚠ Tool output is formatted for LLM consumption; downstream reasoning may be impacted by formatting summaries vs raw details.
- ⚠ Authorization is only as strong as the kubeconfig permissions; over-permissioning increases risk.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for harvester-mcp-server.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-04-04.