F5 Insight Skills

A skill-based automation and troubleshooting platform for F5 BIG-IP and NGINX infrastructure. Combines a chat-driven agent, a visual workflow editor, ECMP autoscaling, and ITSM integration into a single Docker deployment.

What It Does

For network engineers: Run diagnostic and configuration skills against F5 devices from a browser. Skills are defined as Markdown files with YAML blocks — no code to write. Chain skills into automation workflows with drag-and-drop, add approval gates, and trigger them from ServiceNow tickets or Prometheus alerts.

For platform teams: Provide self-service F5 operations with guardrails. Destructive actions require approval. All executions are logged. Credentials never leave the vault. The MCP server lets you plug F5 operations into any AI agent.

Architecture

┌─────────────────────────────────────────────────────────┐
│                    Docker Compose                        │
│                                                         │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌────────┐ │
│  │ Frontend  │  │ Backend  │  │ OpenBao  │  │  MCP   │ │
│  │ React     │→ │ FastAPI  │→ │ Vault    │  │ Server │ │
│  │ :3000     │  │ :8000    │  │ :8200    │  │ :8100  │ │
│  └──────────┘  └────┬─────┘  └──────────┘  └────────┘ │
│                     │                                    │
│  ┌──────────┐  ┌────┴─────┐                             │
│  │Prometheus│→ │Alertmgr  │  (ECMP autoscale alerting)  │
│  │ :9090    │  │ :9093    │                              │
│  └──────────┘  └──────────┘                             │
└─────────────────────────────────────────────────────────┘
         │ SSH / iControl REST
         ▼
   ┌──────────┐  ┌──────────┐
   │ BIG-IP   │  │ NGINX    │
   │ Devices  │  │ Devices  │
   └──────────┘  └──────────┘

Container	Port	Role
insight-frontend	3000	React UI — 10-tab interface
insight-backend	8000	FastAPI — skill execution, automation engine, chat agent
insight-vault	8200	OpenBao — credential storage (auto-init, auto-unseal)
insight-mcp	8100	MCP server — exposes skills as tools for AI agents
insight-prometheus	9090	Metrics collection for autoscale alerting
insight-alertmanager	9093	Alert routing to webhook-triggered automations

Quick Start

git clone https://github.com/YOUR-ORG/f5-insight-skills.git
cd f5-insight-skills

# Configure
cp .env.sample .env
# Edit .env — set ANTHROPIC_API_KEY (or configure OpenAI/local LLM)

# Launch
docker compose up -d --build

# Open
open http://localhost:3000

Vault auto-initializes on first boot. No manual setup required.

UI Tabs

Tab	Purpose
Chat	Natural language interface — describe what you need, the agent picks the right skill
Devices	Add/edit BIG-IP and NGINX devices with SSH or key-based auth
Images	Stage and push BIG-IP ISO images to devices
Topology	Visual virtual server → pool → node topology discovery
Skills	Browse, inspect, and run the 29 built-in skills
Builder	Create new skills with a guided form
Automation	Build and run multi-step skill chains with templates
Visual Editor	Drag-and-drop workflow builder with branching and approval gates
Integrations	Incoming/outgoing webhooks and MCP client connections
History	Chat conversation history

Built-In Skills (29)

Diagnostics (read-only)

bigip-arp-table · bigip-bgp-verify · bigip-boot-locations · bigip-connection-drain · bigip-connection-table · bigip-persistence-records · bigip-pool-status · bigip-route-table · bigip-tcpdump · bigip-topology · bigip-vs-config · nginx-log-analysis · nginx-tcpdump · nginx-upstream-health

Configuration (approval required)

bigip-bgp-withdraw · bigip-config-backup · bigip-config-sync · bigip-fleet-join · bigip-fleet-leave · bigip-irule-install · bigip-irule-remove · bigip-node-toggle · bigip-upgrade · bigip-ve-deprovision · bigip-ve-license · bigip-ve-license-revoke · bigip-ve-provision · bigip-virtual-server-create · bigip-vs-toggle

Automation Templates

Four built-in chain templates for common workflows:

Template	Steps	Use Case
Troubleshoot Connectivity	vs-config → tcpdump → arp-table	Debug traffic flow through a virtual server
Pool Member Maintenance	pool-status → node-toggle (⏸ approval) → connection-table	Safely disable a pool member
ECMP Scale-Out	ve-provision → ve-license → config-sync → bgp-verify → fleet-join (⏸ approval)	Add a BIG-IP VE to the ECMP cluster
ECMP Scale-In	bgp-withdraw (⏸ approval) → connection-drain → ve-license-revoke → fleet-leave → ve-deprovision (⏸ approval)	Gracefully remove a BIG-IP VE

LLM Support

Configure one provider in .env:

Provider	Variables	Notes
Anthropic (default)	ANTHROPIC_API_KEY, ANTHROPIC_MODEL	Claude powers chat agent and skill analysis
OpenAI	OPENAI_API_KEY, OPENAI_MODEL, OPENAI_BASE_URL	GPT-4o or compatible
Local	LOCAL_LLM_BASE_URL, LOCAL_LLM_MODEL	Any OpenAI-compatible endpoint (Ollama, vLLM, llama.cpp)

Without an API key, skills still execute — AI analysis and chat are skipped gracefully.

MCP Server

The MCP server at http://localhost:8100/mcp exposes all skills as tools via the Model Context Protocol. Connect any MCP-compatible client (Claude Desktop, Claude Code, custom agents) to manage F5 infrastructure through natural language.

ServiceNow Mock

A separate ServiceNow-accurate mock is included in snow/ for demonstrating end-to-end ITSM integration. See snow/README.md for setup.

Documentation

Document	Description
DEPLOYMENT-GUIDE.md	Prerequisites, installation, configuration, production hardening
ARCHITECTURE.md	System design, data flow, file structure, extension points
SKILL-AUTHORING.md	How to write custom skills with the SKILL.md format
API-REFERENCE.md	Complete REST API documentation

License

Proprietary — internal use only.