MCP & AI connectors
Secure remote tools that give AI assistants controlled, read-only access to live systems and specialist data — one connector, every tool.
- Streamable HTTP & SSE
- OAuth & token security
- Azure App Service
AI Engineering · MCP & Agents
MCP server development
for Claude, ChatGPT and AI agents.
MCP server development and data engineering in Ireland means production-ready AI, automation and data platforms built for hard business problems—securely, practically and end to end.
Built for real operations
- Model Context Protocol
- Azure OpenAI
- OAuth & governed access
- Python/PostgreSQL
What I do
From messy requirements to
working software.
I help businesses turn ambitious AI and data ideas into secure, maintainable products. One technical owner from discovery and architecture through build, deployment and handover.
AI systems & agents
Useful AI grounded in your business: RAG, intelligent workflows, document review, classification and decision support — including agents that run on a schedule.
- Azure OpenAI & Anthropic
- RAG & knowledge systems
- Scheduled/cron agents
Connected platforms
Full-stack applications and integrations that connect teams, tools and data without creating another silo.
- React & Next.js
- Python & Node.js APIs
- PostgreSQL & webhooks
Selected outcomes
Built for businesses where
the details matter.
AI · Sales operations
Proposals built in minutes, not hours.
An Azure-hosted quotation platform with RAG-powered pricing, proposal generation and direct CRM integration. Pricing runs through a deterministic engine so the AI never invents a number, and every major design decision is written up as an architecture decision record.
Azure OpenAIReactPostgreSQL
View the project ↗
MCP · Financial data
Your broker account, charts and market data inside AI — on a schedule.
A secure MCP connector that lets an AI assistant retrieve broker-account information, analyse market data and create stock charts through a controlled, read-only interface. Every write operation is gated behind a broker-confirmed dry run and a one-use, time-boxed token — and a scheduled agent calls it automatically during market hours to propose entry, exit and hedge decisions.
MCPBroker APIScheduled agent
View the project ↗
MCP · Irish open data
Irish weather and sea conditions, inside AI.
A 28-tool MCP connector that brings weather forecasts, marine warnings, tides, buoys, surf and farming data into Claude and ChatGPT.
MCPMet ÉireannMarine data
View the project ↗
MCP · SEO & analytics
Google Analytics, Search Console and site ops, inside AI.
A 65-tool MCP connector that reads Google Analytics, Search Console and Tag Manager, provisions new sites end to end and manages DNS on Cloudflare, behind a single owner-gated OAuth 2.1 login.
MCPGA4 & GSCCloudflare
View the project ↗
AI · Pharmaceutical validation
A GDPR-aligned RAG platform for regulated document review.
An Azure-hosted RAG platform for pharmaceutical validation documents, built GDPR-compliant from day one — full audit trails, secure document storage and role-based access throughout, hosted entirely inside the client's own Azure tenant.
AzureRAGGDPR
View the project ↗
Client identities and sensitive implementation details are kept confidential.
10+years across data,
finance & technology
finance & technology
About
Senior thinking.
Hands-on delivery.
I'm Sean Twomey, an AI developer and data engineer based in Cork, Ireland, working remotely with UK, US and Irish clients.
My background spans AIB (data lead on a €4bn STS securitisation, including large-scale financial data migration and reconciliation), Bank of America (market-data pipelines) and UnitedHealth Group (Azure/PySpark data migration and source-to-target data mapping across billions of rows)—plus healthcare and pharmaceuticals along the way. That background is where the engineering habits behind my AI work come from: validate before you trust a number, reconcile source against target at every stage, and automate the deployment path with CI/CD pipelines (GitHub Actions, federated/OIDC auth) rather than long-lived secrets.
I carry the same discipline into AI projects: real design decisions get written up as architecture decision records—the options considered, the trade-off, and the risk knowingly accepted—not just the final choice. So what gets built is useful, secure and ready to operate.
More about Sean ↗Based in Cork · Working worldwide
