Configuración MCP & LLM
MCP Console → Dashboard → Configuración General
General
Data Sources
LLM Engine
MCP Servers
Pipeline
Seguridad
Monitoreo
Configuración del MCP Server
⚡
MCP Server Principal
Configuración del orquestador
● RUNNING
Server Name
Host / Port
Transport
Protocolo de comunicación
Max Connections
TLS Enabled
Auto-restart
📐
Context Window & Tokens
Gestión de ventana de contexto
● OK
Max Context
Token Budget
Distribución del contexto
System: 15%Tools: 25%Resources: 35%History: 25%
Truncation
Prompt Caching
Configuración del LLM Engine
🧠
Modelo LLM
Selección y parámetros del modelo
● LOADED
Provider
Model
Modelo activo
API Key
Temperature
Max Tokens Out
Top-P
Streaming
🛡️
Safety & Guardrails
Filtros de seguridad del LLM
● ENFORCED
Prompt Injection
Detección de inyecciones
PII Detection
Datos personales
Hallucination Guard
Content Filter
Max Tool Loops
Límite de iteraciones
Timeout (ms)
Audit All Calls
📝
System Prompt
Instrucciones base para el LLM conectado al MCP
● ACTIVE
Data Sources Conectados
🗄️
Data Sources Registrados
Fuentes de datos conectadas al MCP
16 ACTIVE
🗄️
PostgreSQL — Core DB
RDBMS
Connected
📊
Snowflake — Analytics DWH
DWH
Connected
🔗
Salesforce CRM
SaaS API
Connected
💰
SAP S/4HANA Finance
ERP
Connected
📈
Google Analytics 4
Analytics
Connected
🔄
Apache Kafka
Stream
Connected
📧
HubSpot Marketing
SaaS API
Rate limited
📦
MongoDB — Products
NoSQL
Connected
MCP Servers & Tools Registrados
🗄️
DB Query Server
mcp-server-db
● RUNNING
Transport
stdio
Tools
query_sql list_tables describe_schema
Resources
db://schemas
Latency
avg 22ms
📊
Analytics Server
mcp-server-analytics
● RUNNING
Transport
SSE
Tools
run_report get_kpi forecast
Resources
analytics://dashboards
Latency
avg 45ms
💰
Finance Server
mcp-server-finance
● RUNNING
Transport
SSE
Tools
get_balance reconcile fx_rate
Resources
finance://ledger
Latency
avg 18ms
🔍
Search Server
mcp-server-search
● RUNNING
Transport
stdio
Tools
search_docs semantic_search
Resources
search://index
Latency
avg 35ms
🔔
Alerts Server
mcp-server-alerts
● RUNNING
Transport
WebSocket
Tools
send_alert get_rules check_anomaly
Channels
Slack Email PagerDuty
Latency
avg 8ms
📄
Docs Server
mcp-server-docs
● RUNNING
Transport
stdio
Tools
read_doc gen_report export_pdf
Sources
GDrive SharePoint
Latency
avg 95ms
Monitoreo en Tiempo Real
📈
Pipeline Health
Estado del pipeline en vivo
● HEALTHY
Requests/min
2,847
Avg Latency
38ms
Error Rate
0.03%
Tool Calls/min
1,204
Cache Hit Rate
87.4%
📋
Activity Log
Últimos eventos del sistema
● STREAMING
12:04:32 [OK] MCP Server heartbeat: all 6 servers healthy
12:04:30 [LLM] Inference: claude-sonnet-4-5 · 42K ctx · 2 tool calls · 380ms
12:04:28 [WARN] HubSpot API rate limit approaching: 82% of quota
12:04:25 [OK] Tool: query_sql executed on PostgreSQL · 22ms · 847 rows
12:04:22 [MCP] Context assembled: system + 4 tools + 2 resources = 38K tokens
12:04:20 [OK] Safety check passed: no PII detected, no injection attempts
12:04:18 [OK] Alert sent: Budget overrun Marketing → Slack #finance-alerts
12:04:15 [AUDIT] User: CFO · Query: "Q3 revenue by region" · Response: 280ms