SuperAPI AI Directory

telemetry-ops

General & other

View repository

Updated 10-15 17:21

About this skill

Unified Telemetry Operations Skill

Comprehensive network telemetry and event collection across multiple protocols.

Skill ID

telemetry-ops

Description

This meta-skill provides a unified interface to all NetClaw telemetry receivers: syslog, SNMP traps, IPFIX/NetFlow, and gNMI streaming telemetry. It enables holistic network monitoring by aggregating events from multiple sources.

When to Use

  • Setting up comprehensive network monitoring across multiple telemetry types
  • Correlating events across syslog, SNMP traps, and flow data
  • Investigating network issues using multiple data sources
  • Understanding the full picture of network health and behavior
  • Onboarding a new device to NetClaw monitoring

Component Skills

Skill MCP Server Protocol Default Port
syslog-receiver syslog-mcp RFC 5424/3164 UDP 514
snmptrap-receiver snmptrap-mcp SNMPv1/v2c/v3 UDP 162
ipfix-receiver ipfix-mcp IPFIX/NetFlow UDP 2055
gnmi-telemetry gnmi-mcp gNMI gRPC 57400

Example Workflows

Full Device Onboarding

1. Configure device to send syslog to NetClaw (UDP 514)
2. Configure SNMP traps to NetClaw (UDP 162)
3. Configure NetFlow/IPFIX export to NetClaw (UDP 2055)
4. Add device to gNMI targets for streaming telemetry
5. Start all receivers
6. Verify data is being received from each source

Multi-Source Incident Investigation

1. Query syslog for error messages around incident time
2. Check SNMP traps for linkDown events
3. Analyze flows for traffic anomalies
4. Subscribe to gNMI telemetry for real-time interface state

Network Health Dashboard

1. Use syslog_get_severity_counts for error distribution
2. Use snmptrap_get_counts for trap type breakdown
3. Use ipfix_top_talkers for bandwidth consumers
4. Use gnmi_get for current device state

Sample Prompts

  • "Start all telemetry receivers on their default ports"
  • "What events have we received from 192.168.1.1 across all sources?"
  • "Show me a summary of network health from all telemetry"
  • "Configure the Catalyst 9300 for full telemetry to NetClaw"
  • "Investigate the network issue at 3pm - check all telemetry sources"

Cisco Catalyst 9300 Configuration

Syslog

logging host 10.0.0.1 transport udp port 514
logging trap informational
logging source-interface Loopback0

SNMP Traps

snmp-server enable traps
snmp-server host 10.0.0.1 version 2c public

NetFlow/IPFIX

flow exporter NETCLAW
 destination 10.0.0.1
 transport udp 2055
 export-protocol ipfix

gNMI

netconf-yang
gnmi-yang
gnmi-yang secure-server

Remote Access (UDP Tunneling)

Since ngrok doesn't support UDP, use these alternatives:

Service UDP Support Best For
Pinggy Yes Quick tunnel setup
Tailscale Yes Persistent mesh VPN
LocalXpose Yes Full protocol support

Architecture

                    ┌─────────────────┐
                    │ Cisco Cat 9300  │
                    └────────┬────────┘
                             │
        ┌────────────────────┼────────────────────┐
        │         │          │          │         │
        ▼         ▼          ▼          ▼         │
   ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐
   │ Syslog  │ │  SNMP   │ │  IPFIX  │ │  gNMI   │
   │ UDP 514 │ │ UDP 162 │ │UDP 2055 │ │TCP 57400│
   └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘
        │           │           │           │
        ▼           ▼           ▼           ▼
   ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐
   │syslog-  │ │snmptrap-│ │ ipfix-  │ │ gnmi-   │
   │  mcp    │ │   mcp   │ │   mcp   │ │   mcp   │
   └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘
        │           │           │           │
        └───────────┴─────┬─────┴───────────┘
                          │
                    ┌─────┴─────┐
                    │  NetClaw  │
                    │   Agent   │
                    └───────────┘

Limitations

  • All receivers use in-memory storage (data lost on restart)
  • No cross-source correlation built-in (done by agent)
  • Each receiver runs independently
  • UDP tunneling required for remote testing

Related Documentation

  • /mcp-servers/syslog-mcp/README.md
  • /mcp-servers/snmptrap-mcp/README.md
  • /mcp-servers/ipfix-mcp/README.md
  • /mcp-servers/gnmi-mcp/README.md

Failure Behavior

  • On a tool error (timeout, unreachable host, malformed response), report the failure and its error message directly to the user rather than fabricating or guessing at results.
  • For a confirmed read-only call, check connectivity and retry once if appropriate. For any call that changes state or sends a message, a timeout does not prove the action failed: inspect current state or delivery status before retrying, preserve the required approval/change gates, and do not repeat an action whose outcome is unknown.