Making AI Personal: Building Context-Rich AI Tools with MCP • Kris Jenkins • GOTO 2025

2026-05-29 · Source: GOTO Conferences · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, extended

Summary

Kris Jenkins' GOTO 2025 talk, "Making AI Personal: Building Context-Rich AI Tools with MCP," advocates for "thinking small" in AI development, focusing on immediate, personal productivity gains rather than large-scale enterprise solutions. He introduces "agentic AI" as a Large Language Model (LLM) augmented with programmer-defined tools, where the LLM intelligently decides when to invoke these tools to enhance context and answer ambiguous questions. The core technology is the Model Context Protocol (MCP), described as JSON RPC, which enables defining these tools and their invocation context. Jenkins provides a practical guide for building an MCP server, requiring only three basic JSON messages: initialize, list your tools, and call a tool. He emphasizes using reflection in programming languages to convert function signatures into JSON descriptions, crucial for the LLM's understanding. The talk also highlights "delegated coding" as an effective prompting strategy for LLMs to assist in server development and stresses the importance of detailed function descriptions for effective tool invocation.

Key takeaway

For AI Engineers or developers seeking to personalize AI for daily tasks, consider building a Model Context Protocol (MCP) server. This allows you to integrate custom tools with LLMs, transforming ambiguous personal questions into actionable insights. You can automate routine tasks like status updates or secure email management by defining specific functions and their invocation context. Start by implementing a basic JSON RPC server with reflection to quickly create context-rich AI agents for your immediate needs.

Key insights

Agentic AI combines LLMs with programmer-defined tools, enabling context-rich answers to ambiguous personal queries.

Principles

• Agentic AI = LLM + tools.
• Programmers define tools, LLM invokes them.
• Documentation is critical for LLM tool use.

Method

Build an MCP server using JSON RPC with initialize, list tools, and call function messages. Use reflection to convert function signatures to JSON descriptions for LLM context.

In practice

• Automate weekly status updates.
• Securely manage emails with narrow access.
• Customize library search criteria.

Topics

• Agentic AI
• Model Context Protocol
• LLM Tooling
• JSON RPC
• Delegated Coding
• Personal Productivity

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• How AI Agents Use MCPs to Deliver Stronger Results Than Standalone LLMs — AI Agents leveraging Model Context Protocol (MCP) transform LLMs from conversational assistants into action-oriented problem solvers by integrating external systems.
• FastMCP: The Pythonic Way to Build MCP Servers and Clients — FastMCP simplifies Model Context Protocol (MCP) server and client development using Python decorators and async support.
• From Tool Calling to MCP: Building a Natural Language Search with Spring AI and MCP Server — MCP standardizes AI model communication with external tools, enabling reusability and independent deployment.
• MCP-Persona: Benchmarking LLM Agents on Real-World Personal Applications via Environment Simulation — MCP-Persona benchmarks LLM agents' significant struggles with personalized tool use in real-world social applications, highlighting a critical evaluation gap.
• Whitepaper Companion Podcast - Agent Tools & Interoperability with MCP — MCP standardizes LLM tool integration, enabling real-world actions and fostering a modular, scalable AI agent ecosystem.
• How to Build a Personal Context MCP — A Personal Context Portfolio (PCP) standardizes user information for AI agents, reducing context repetition and product lock-in.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by GOTO Conferences.