Notes on agent systems, workflow automation, and practical AI usage in engineering work.
Build a Hugging Face AI Stack With MCP and Python Clients: An infrastructure-oriented example that uses Hugging Face models behind an MCP server and calls task-specific models from a Python MCP client.
Build an MCP Server in Python: A reference example for building a simple MCP server in Python with the official MCP Python SDK.
Call an MCP Server From a Python Service: A follow-up example showing how a Python service can connect to an MCP server and call its tools over streamable HTTP.
Organize AI Workflows as Files, Not Frameworks: A file-oriented AI architecture pattern where one agent connects to multiple workflows, and each workflow is broken into tasks, prompts, data, and tools.
Prompt Engineering, Context Engineering, and Harness Engineering: A practical distinction between prompt engineering, context engineering, and harness engineering, and why all three matter when moving from demos to reliable AI systems.
Salesforce AgentForce: A quick reference on Salesforce Agentforce, its Atlas reasoning engine, and how autonomous agent flows are structured.