Please Make Jarvis
(so I don't have to)
Let's wake up
Your brain needs this 🧠
What's a Jarvis?
Generated UI?
Let's build Jarvis
Seamless AI Assistance
- Understands the user
- Take Action
- Across All Services
- Any Device, Anywhere
The Paradigm Shift
Today: Browsers & Websites
Future: Jarvis Clients & MCP Services
The Building Blocks
- The Jarvis Client
- Service Auth (OAuth 2)
- Task Orchestration
- Security & Trust
- Personalization & Memory
- Interconnected MCP Services (lots of them)
- One more thing...
The Bottleneck:
Foundation & Friction
Foundation: MCP
- The Standard API for AI Agents
- Defines Service Capabilities
- Enables Agent ↔️ Service Communication
Friction: Tool Overload
- Jarvis needs MCP Servers
- But... Thousands? Millions?
- LLMs get confused with too many options!
Inspiration: How Humans Cope
- Search (Google)
- Experience & Memory
- Bookmarks & Favorites
The Solution:
Jarvis Search Engine
The Search Engine for MCPs
- Finds relevant MCP services based on intent
- Provides a ranked list
Smarter than PageRank
- User History
- Task Success Rate
- Context & Relevance
- Personalization
Let's build Jarvis' Search Engine
You can do it!
Thank you!
𝕏 @kentcdodds
Please make Jarvis (so I don't have to)
By Kent C. Dodds
Please make Jarvis (so I don't have to)
The dream of a seamless AI assistant like Tony Stark's Jarvis – capable of understanding intent and accomplishing tasks across any service, anywhere – is closer than ever, but key infrastructure is missing. While the Model Context Protocol (MCP) provides a crucial foundation for standardized agent-service communication, replacing websites with MCP services, it introduces a critical challenge: tool overloading. How can an AI agent effectively choose the right service when faced with potentially millions of options? This talk outlines the requirements for building a true "Jarvis." We need a "Google for MCPs"—an intelligent discovery and ranking engine that leverages not just service descriptions, but user history, task success rates, and deep personalization to overcome tool overload and connect users' intentions to the right MCP service. Join this session to explore the architecture required for this future and understand the call to action: let's build the necessary infrastructure to make Jarvis a reality.
