esxi-readonly-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| ESXI_HOST | Yes | IP o nombre del ESXi (o del vCenter) | |
| ESXI_PORT | No | Puerto de la API | 443 |
| ESXI_USER | Yes | Usuario de solo lectura | |
| ESXI_PERF_DB | No | Ruta de la base SQLite del historial de performance | ~/.esxi-readonly-mcp/perf_historial.db |
| ESXI_PASSWORD | No | Solo si no podés usar keyring (queda en texto plano en la config: evitalo) | |
| ESXI_VERIFY_SSL | No | 1 para verificar el certificado (por defecto se acepta el autofirmado de ESXi) | 0 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| verificar_permisosA | Muestra con que usuario se conecta y si ese usuario tiene privilegios de escritura en ESXi. |
| host_infoB | Host: fabricante, modelo, CPU (cores/hilos/HT), RAM total y usada, version y build de ESXi, uptime, politica de energia y controladoras de storage. |
| datastoresB | Datastores: capacidad, usado, libre, % de uso, % provisionado (thin), disco fisico donde esta cada uno y VMs. |
| vmsC | Todas las VMs: estado, vCPU, RAM asignada/activa/consumida, ballooning, swap, reservas/limites, VMware Tools, discos (provisionado vs usado, thin/thick, datastore, controladora) y espacio libre dentro del guest. |
| snapshotsA | Todos los snapshots con fecha, antiguedad, tamano aproximado y marca de los que superan dias_alerta. |
| performanceA | Metricas de host y VMs: CPU uso, CPU Ready %, Co-Stop %, memoria activa/ballooning/swap, latencia de disco (total, por disco virtual y por datastore) e IOPS. Devuelve promedio, p95 y maximo con su momento. intervalo: realtime (20 s, ~1 h en ESXi standalone), dia, semana, mes, anio (estos requieren vCenter). max_muestras: 180 = 1 h en realtime. dias: rango hacia atras (alternativa a max_muestras). |
| historial_perfC | Resumen de las metricas guardadas localmente por 'server.py --collect' (semanas de historia, aunque ESXi solo guarde 1 h). Filtra por horario laboral (lunes a viernes, hora local). |
| top_archivosB | Archivos mas grandes de cada datastore (o de uno), con la VM a la que pertenecen y marca de posibles huerfanos. |
| overcommitB | Suma de vCPU vs cores/hilos fisicos y RAM asignada vs fisica (ratios de overcommit). |
| salud_hardwareA | Salud del hardware segun ESXi: estado de la RAID y discos fisicos (PERC), sensores de temperatura, ventiladores, fuentes, memoria y CPU. Lista primero los elementos que no estan en verde. |
| eventosB | Eventos de ESXi (errores de disco, reinicios de VMs, alarmas, tareas, logins fallidos). nivel: todos | warning (warning y error) | error. Fechas corregidas por el desfase del reloj de ESXi. |
| config_hostC | Configuracion del host: licencia (clave enmascarada) y vencimiento, NTP y desfase del reloj, SSH/Shell, servicios activos, perfil de imagen y VIBs/drivers de terceros, syslog, placas de red (velocidad de enlace), vSwitches, portgroups y VMkernel. |
| diagnostico_completoB | Ejecuta todo junto: permisos, host, salud de hardware, configuracion, datastores, VMs, snapshots, performance realtime, eventos (warning/error 7 dias), top archivos y overcommit. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 13 tools
Each tool targets a distinct ESXi monitoring area: permissions, host hardware, datastores, VMs, snapshots, performance, historical performance, largest files, overcommit, hardware health, events, host config, and a full diagnostic aggregator. Overlaps are minor and well explained, especially performance vs. historial_perf and the all-in-one diagnostico_completo. An agent can reliably select the right tool for a specific monitoring question.
All names use lowercase snake_case, but they mix English and Spanish terms (e.g., host_info, datastores, eventos, historial_perf) and mix noun phrases with one verb phrase (verificar_permisos). The pattern is readable but not predictable or consistent in language or grammatical style. A more uniform convention would improve clarity.
With 13 tools, the server is well-scoped for read-only ESXi diagnostics. Each tool covers a meaningful monitoring concern, and the aggregate diagnostico_completo tool provides a convenient full-scan option without bloating the set. This is within the ideal 3-15 range for an MCP server.
Coverage is broad: host, hardware health, config, datastores, VMs, snapshots, performance, events, overcommit, top files, and permissions are all represented. Minor gaps exist, such as explicit per-VM network details and local user/role enumeration beyond the current connection. These gaps are workable for most read-only diagnostic workflows.