Model Context Protocol moderate risk

Last9

Last9 MCP server provides AI agents with real-time production context (logs, metrics, traces) to assist in code troubleshooting and automated fixes.

Connections & Capabilities

Connects To

GitHubKubernetesAWS

Capabilities

readwrite

Quickstart

Install

brew install last9-mcp

Config

{
  "mcpServers": {
    "last9": {
      "command": "/opt/homebrew/bin/last9-mcp",
      "env": {
        "LAST9_REFRESH_TOKEN": "<last9_refresh_token>",
        "OTEL_EXPORTER_OTLP_ENDPOINT": "<otel_endpoint_url>",
        "OTEL_EXPORTER_OTLP_HEADERS": "<otel_headers>"
      }
    }
  }
}

Exposed MCP Tools (21)

safe

get_exceptions

Retrieves server-side exceptions within a specified time range.

Read-only operation, retrieves existing data.

safe

get_service_summary

Provides a summary of a service's performance metrics (throughput, error rate, response time) over a given time range.

Read-only operation, retrieves aggregated data.

safe

get_service_environments

Lists available environments for services.

Read-only operation, retrieves environment names.

safe

get_service_performance_details

Retrieves detailed performance metrics for a specific service.

Read-only operation, retrieves performance data.

safe

get_service_operations_summary

Provides a summary of operations within a service, such as HTTP endpoints and database queries.

Read-only operation, retrieves operation summaries.

safe

get_service_dependency_graph

Retrieves a dependency graph showing incoming and outgoing dependencies for a service.

Read-only operation, retrieves dependency information.

safe

prometheus_range_query

Executes a PromQL range query to retrieve metrics data over a specified time range.

Read-only operation, retrieves metrics data.

safe

prometheus_instant_query

Executes a PromQL instant query to retrieve metrics data at a specific point in time.

Read-only operation, retrieves metrics data.

safe

prometheus_label_values

Returns label values for a given label and PromQL filter query.

Read-only operation, retrieves label values.

safe

prometheus_labels

Returns labels for a given PromQL match query.

Read-only operation, retrieves labels.

safe

get_logs

Retrieves logs filtered by service name and/or severity level.

Read-only operation, retrieves log data.

safe

get_drop_rules

Retrieves drop rules for logs, which determine what logs are filtered out.

Read-only operation, retrieves existing drop rules.

moderate

add_drop_rule

Adds a new drop rule to filter out specific logs.

Allows modification of log filtering, potentially hiding critical information.

safe

get_service_logs

Retrieves raw log entries for a specific service over a time range.

Read-only operation, retrieves log data.

safe

get_log_attributes

Retrieves available log attributes (labels) for a specified time window.

Read-only operation, retrieves log attributes.

safe

get_traces

Retrieves traces using JSON pipeline queries for advanced filtering.

Read-only, retrieves trace data.

safe

get_service_traces

Retrieves traces by trace ID or service name with time range filtering.

Read-only, retrieves trace data.

safe

get_trace_attributes

Retrieves available trace attributes (series) for a specified time window.

Read-only, retrieves trace attributes.

safe

get_change_events

Retrieves change events from the last9_change_events Prometheus metric.

Read-only, retrieves change event data.

safe

get_alert_config

Retrieves alert configurations (alert rules) from Last9.

Read-only, retrieves alert configurations.

safe

get_alerts

Retrieves currently active alerts from the Last9 monitoring system.

Read-only, retrieves active alert data.

Safety Assessment

The Last9 MCP server is generally safe for read-only operations. The `add_drop_rule` tool introduces some risk, requiring careful configuration and monitoring. Prometheus queries should be crafted with caution to avoid performance impacts.

Most tools are read-only, retrieving observability data.
Write operations are limited to adding log drop rules.
No direct code execution or system-level access.
Clear tool descriptions and parameter definitions.

The `add_drop_rule` tool allows modification of log filtering, potentially hiding critical information if misconfigured.
Prometheus queries can potentially overload the system if not carefully crafted.
Lack of explicit sandboxing for Prometheus queries.
Incorrectly configured alert rules could lead to alert fatigue or missed critical issues.