ai-integration-architect

AI Integration Architect — Claude Code Plugin

A Claude Code plugin that helps teams plan, design, and scaffold AI integration into enterprise systems.

46% of enterprises say integrating AI into existing systems is their #1 challenge. This plugin acts as a senior integration architect — it assesses your system landscape, recommends architecture patterns, generates production-ready code scaffolds, and produces deployment plans with security baked in.

Installation

Option 1: One-command install via marketplace (recommended)

claude plugin marketplace add pfbarros2/ai-integration-architect-plugin
claude plugin install ai-integration-architect

Option 2: Install directly from the skill repo

claude plugin marketplace add pfbarros2/ai-integration-architect
claude plugin install ai-integration-architect

Option 3: Local install

git clone https://github.com/pfbarros2/ai-integration-architect-plugin.git
claude --plugin-dir ./ai-integration-architect-plugin

What it does

Phase	Output
Assess	Maps your current systems (APIs, databases, SaaS tools) and identifies where AI adds the most value
Architect	Recommends integration patterns (MCP servers, API gateways, RAG pipelines, event-driven agents) with trade-off analysis
Scaffold	Generates working starter code — MCP servers, API connectors, RAG pipelines, middleware — with auth, rate limiting, error handling, and tests
Deploy	Produces deployment configs, monitoring setup, security checklists, cost estimates, and phased rollout plans

Example prompts

Once installed, just describe your integration challenge:

"We run Shopify Plus with Salesforce CRM and Zendesk for support. Our CEO wants AI to help reduce support response times. Where do I start?"

"I need an MCP server that connects Claude to our internal REST API. It uses OAuth2, has /employees, /projects, and /timesheets endpoints."

"We have 5000 pages of docs across Confluence, Google Docs, and Notion. Engineers waste hours searching. I want a RAG pipeline so our AI assistant can answer questions from any internal doc."

"Design an event-driven architecture where an AI agent triages incoming support tickets in Zendesk, looks up customer history in Salesforce, and drafts responses."

Supported integration patterns

• MCP Servers — Connect Claude to any API, database, or internal tool (TypeScript + Python templates)
• API Gateways — Expose AI as a service for your existing systems to call, with model routing (Haiku/Sonnet/Opus by complexity)
• RAG Pipelines — Ingest enterprise knowledge from Confluence, SharePoint, Notion, Google Docs; chunk, embed, retrieve with reranking
• Event-Driven Agents — AI that responds to system events (new ticket, alert, deployment) with human-in-the-loop approval gates
• Hybrid Architectures — Combine patterns for complex workflows

Security built in

Every output includes enterprise security by default:

• Least-privilege service accounts and scoped API keys
• Audit logging for all AI-initiated actions
• Data classification guidance (Public → Internal → Confidential → Restricted)
• Compliance checklists for SOC 2, GDPR, and HIPAA
• Prompt injection mitigation strategies

Plugin structure

ai-integration-architect-plugin/
├── .claude-plugin/
│   └── plugin.json
├── skills/
│   └── ai-integration-architect/
│       ├── SKILL.md
│       └── references/
│           ├── patterns.md
│           ├── scaffolds.md
│           ├── security.md
│           └── deployment.md
├── marketplace.json
└── README.md

License

MIT