Well-designed, documented, and versioned interfaces are the foundation for connecting AI systems to enterprise data and business functions. Building on this, I design MCP-based integrations, agent workflows, and production-ready AI architectures that connect LLMs to tools, resources, and APIs in a secure, observable, and operable way. The focus is not on isolated demos, but on real systems: structured tool access, orchestration, governance, and the engineering patterns required to move from copilots to reliable AI automation.
What Friedrichs IT delivers
Production-ready AI integrations with MCP, orchestration, observability, and governance — designed to connect LLMs to real systems in a secure, traceable, and operable way.
MCP Servers
I develop robust MCP servers that make existing APIs and data sources usable for LLM workflows. This includes tracking changes to the MCP schema, selecting the right transport for each use case, and designing tools, resources, and prompts to be token-efficient, reliable, and testable. OAuth-based security, proper secrets handling, and production-ready server operation are built in from the start.
LightNow
I am the founder of LightNow, a platform for the secure use of MCP servers in enterprises. LightNow combines trust, capabilities, policies, and usage: capabilities and trust signals are visible per server and version, policies define which servers are allowed, and installation and actual usage become easier to govern across teams and environments.
Agent Workflows
I design and implement AI workflows that combine LLMs with tools, APIs, and business logic. This includes MCP client integration, orchestration with platforms and frameworks such as n8n, LangChain, and LangGraph, and the connection of chatbots or assistants to real systems. The result is not just a chatbot, but an operable workflow that can retrieve context, call tools, and execute defined tasks across multiple steps.
Observability & Evaluation
Production AI needs more than prompts and tool calls. I add tracing, evaluation, and operational visibility for multi-step AI workflows, including tool usage, latency, failures, and decision paths. Observability stacks such as Langfuse or Phoenix help make agent behavior debuggable, measurable, and safer to operate in real environments.
How I can help
- Build MCP-based AI integrations for APIs, tools, and internal systems.
- Design agent workflows with clear orchestration and execution boundaries.
- Add tracing, evaluation, and observability to production AI systems.
- Establish governance, trust, and controlled rollout for enterprise AI.
From Expertise to Insights