MCPI is an open protocol enabling AI agents to discover, verify, and transact with web services through a trust-based economic network with efficient agent-website introduction.
MCPI extends the Model Context Protocol to create a bridge between AI and web services.
Find web services through DNS-based discovery with standardized capability descriptions.
Efficient introduction between AI agents and websites with context-aware customization and capability advertising.
Operate in active mode as a middleware or passive mode as a directory service.
Enable AI agents to design and execute complex solutions by combining specialized services across vendor relationships.
Seamlessly integrate with existing websites and services using standard web technologies and protocols.
Extend capabilities through modular plugins that can be loaded and configured at runtime.
The web is at an inflection point as AI agents become the primary interface between humans and digital services.
We're entering the most profound transformation of the internet since its inception. After 30 years of humans interacting directly with web interfaces, AI agents will increasingly mediate our digital experiences. Just as early standards like HTTP and HTML shaped the web, the protocols emerging now will define AI-web interactions for decades to come.
MCPI extends the Model Context Protocol to create a universally accessible standard for AI agents to discover, verify, and transact with web services through DNS-based discovery.
Advantage: Open, efficient, and complementary to all agent ecosystems. MCPI works alongside existing solutions while ensuring the web remains accessible to all.
History has consistently shown that open standards eventually win against proprietary alternatives. The foundation of the internet is built on open protocols that anyone can implement.
Just as roads optimized for cars are different from pedestrian walkways, interfaces designed for AI agents should be different from human UIs. MCPI enables 90%+ reduction in token usage.
By leveraging DNS TXT records, MCPI enables agents to discover services through the existing internet infrastructure, creating a seamless bridge between the current web and the emerging agent network.
MCPI enables AI agents to orchestrate complex solutions across vendor boundaries, unlocking unprecedented collaboration that would be impractical in closed ecosystems.
A new era of accessible AI sparks widespread interest in AI-human interactions through conversation.
AI systems navigate the human web with significant token usage and efficiency challenges.
Various approaches emerge for how AI agents should interact with digital services.
MCPI introduces an open, DNS-based protocol that creates a universal, vendor-neutral standard for agent-service discovery and interaction.
Current approaches to AI-web connectivity often focus on having AI agents browse human-designed websites, creating specialized agent interfaces, or building proprietary infrastructure. While these approaches have merits, they risk creating closed ecosystems that limit innovation and interoperability.
MCPI offers an open alternative that works alongside these approaches, ensuring the emerging agent-powered web remains as open and accessible as its predecessors.
The Hello Protocol extension enables AI agents to efficiently understand websites without requiring token-intensive HTML parsing.
Currently, AI agents must ingest and parse entire websites to understand what services they offer—a process that is costly, inefficient, and often results in incomplete understanding. The Hello Protocol solves this by providing a standardized way for websites to introduce themselves to AI agents with customizable, context-aware responses.
AI agent discovers an MCPI-enabled website through DNS
Agent sends a HELLO operation request with optional context
Website responds with tailored introduction and capability information
Agent now efficiently understands the website's purpose and capabilities
// Hello Protocol Request Example
{
"jsonrpc": "2.0",
"id": "request-id",
"method": "tools/call",
"params": {
"name": "website_content",
"arguments": {
"operation": "HELLO",
"context": "shopping",
"detail_level": "standard"
}
}
}
// Hello Protocol Response Example
{
"jsonrpc": "2.0",
"id": "request-id",
"result": {
"content": [
{
"type": "text",
"text": "Welcome to Example Store! We offer a wide range of eco-friendly products with free shipping on orders over $50. Our most popular categories include kitchen supplies, personal care, and home goods."
}
],
"metadata": {
"provider": {
"name": "Example Store",
"domain": "example.com",
"description": "Online retailer of eco-friendly products"
},
"capabilities": [
"product_search",
"customer_lookup",
"order_history"
],
"topics": [
"sustainability",
"eco-friendly",
"zero-waste"
]
}
}
}
Reduces token usage by 90%+ compared to HTML parsing
Customize introductions based on visitor intent
Optimize for AI agent discovery with tailored content
Immediately inform AI agents about available tools
How MCPI expands MCP from developer tools to the entire web ecosystem
Currently, Model Context Protocol (MCP) is primarily used by developers to connect AI systems to development tools:
"MCP has revolutionized how AI interfaces with development environments, but its potential extends far beyond coding tools."
MCPI expands this vision to become a universal gateway between AI and all web services:
"MCPI transforms MCP from a developer-focused tool into a universal gateway for AI to interact with the entire web on behalf of users, businesses, and services."
"Just as roads are built for cars, not pedestrians, MCPI creates a web infrastructure designed specifically for AI agents rather than human browsers."
By building on MCP's foundation, MCPI creates a bridge that enables AI systems to:
Direct capability access eliminates the need for slow HTML parsing, content extraction, and page navigation that traditional web browsing requires
Structured data reduces token usage by 90%+ compared to parsing HTML, while reverse fee economics shift costs from users to service providers
Purpose-built for AI interaction rather than human visual consumption, enabling complex capabilities impossible with traditional web browsing
Execute complex business processes across multiple services that would be impractical or impossible via traditional web browsing
MCPI enables a progression toward a truly intelligent web:
How MCPI enables AI agents to create dynamic solution packages across vendor relationships
One of MCPI's most powerful capabilities is enabling AI agents to design and execute complex solutions by dynamically orchestrating multiple specialized services from different vendors. The relationship network is key to this process, allowing AIs to identify compatible services with established trust connections.
The Concierge Agent analyzes complex requests, designs multi-vendor solutions using the trust network, and coordinates execution across specialized service providers. This orchestration happens behind the scenes, presenting users with a seamless solution while handling all the complexity of multi-vendor coordination.
Creates a complete manufacturing solution by bringing together design engineering, precision machining, advanced materials sourcing, and electronic component integration with a unified production plan where each specialized vendor contributes their expertise to a cohesive product.
Builds a comprehensive marketing solution that combines market research, creative development, content production, media distribution, and performance analytics into a unified campaign architecture where specialized vendors work in concert through shared data and assets.
Creates a tailored business technology solution by combining core systems, custom development, data services, and specialized integrations into a technical architecture where multiple vendors' components work together seamlessly as a complete transformation package.
These multi-vendor solutions are orchestrated by Concierge Agents that leverage the MCPI protocol's trust relationships to coordinate specialized services seamlessly.
How brands, users, and designers will interact when AI agents become the primary interface to web services
As AI agents become the primary way we interact with web services, we're entering a new era where content will be primarily accessed through AI interfaces rather than direct browsing. But this doesn't mean the end of human expression or brand identity - just a transformation in how they're conveyed.
Users might say "show me this data as a minimalist dashboard" or "present this in a playful style" and the AI would render the information according to preferences, accessibility needs, or context.
Example: "Claude, show me the weather forecast as an animated children's book for my kids"
Brands will maintain their identity through MCPI brand definitions that AI agents can interpret when rendering content to users.
"branding": {
"colors": {
"primary": "#3498db",
"secondary": "#2ecc71"
},
"logo": {
"vector": "https://example.com/logo.svg"
},
"typography": {
"primary": "Helvetica Neue"
},
"tone": "professional"
}
AI interfaces might offer "view original" options, while services could provide both structured data for AI consumption and rendering hints for human display.
Example: An AI might say "I've summarized this news article, but you can see the publisher's original format if you prefer"
Instead of designing pages, companies will create brand experience guidelines for AI rendering. Design systems might include "AI voice/personality" alongside visual components.
Example: Brand guidelines evolving to include "Interaction Principles" and "AI Conversation Patterns"
This evolution creates an interesting middle ground where users get personalized presentations of information while brands maintain recognizable identity elements. MCPI can include brand definition components, allowing services to express how they'd prefer to be represented while still giving users flexibility in how they consume the information.
It's similar to how voice interfaces for brands work today - companies define their voice assistant personality while the actual rendering happens on the user's device in a way that makes sense for that medium.
Our phased approach to developing and standardizing the open MCPI protocol specification.
The development roadmap for the community-driven MCPI protocol specification
Building the core protocol specification and reference implementation.
Bootstrapping the initial network and building adapters.
Implementing the economic framework and payment infrastructure.
Scaling the network with advanced features and governance.
A glimpse of how MCPI works with DNS discovery and capability execution.
// DNS TXT Record with Multiple Endpoints for Redundancy
_mcp.example.com IN TXT "v=mcp1 endpoints=https://mcp1.example.com/api,https://mcp2.example.com/api capabilities=product_search,customer_lookup"
// Discovery Request
{
"method": "mcpi/discover",
"params": {
"detail": "standard"
}
}
// Discovery Response
{
"provider": {
"name": "Example Store",
"domain": "example.com",
"description": "Online retailer of eco-friendly products"
},
"mode": "active",
"capabilities": [
{
"name": "product_search",
"description": "Search for products in catalog",
"category": "inventory",
"operations": ["SEARCH", "GET"]
}
],
"referrals": [
{
"name": "Eco Shipping",
"domain": "ecoshipping.com",
"relationship": "trusted"
}
]
}
Comprehensive guides to help you implement and use the MCPI protocol.
Learn how MCPI is reshaping the web for AI agents through practical examples and workflows.
MCPI provides fundamental advantages over having AI agents browse websites designed for humans:
Traditional web browsing by AI agents will always be slower, more expensive, less reliable, and more limited than direct MCPI integration, similar to how APIs outperform screen scraping for developers.
The Infrastructure Analogy: Just as roads are built specifically for cars rather than pedestrians, MCPI creates an infrastructure layer designed specifically for AI agents rather than human visitors. Asking AI to browse human-designed websites is like asking cars to drive on sidewalks — possible but fundamentally inefficient and limiting.
The dual-mode operation enables flexible integration models:
A retail platform operates in active mode, directly connecting to inventory, payment, and shipping systems. AI agents interact only with the retail platform, which handles all backend operations.
A medical directory operates in passive mode, providing AI agents with verified referrals to specialized healthcare providers based on needed capabilities, with providers handling their own operations.
MCPI is the foundational layer in a three-stage evolution toward a more intelligent web:
The base infrastructure layer that enables efficient AI-website connections, service discovery, and standardized interactions.
Domain-specific intelligent representatives that serve as the primary interface for a digital service, coordinating capabilities and providing personalized interactions.
Interconnected mesh networks of specialized intelligent agents that can discover, communicate with, and delegate tasks to each other, forming dynamic teams to solve complex problems.
Each stage builds on the previous one, creating a progression toward increasingly sophisticated AI-powered web interactions.
MCPI enables AI agents to dynamically compose specialized services from multiple vendors into cohesive solution packages:
The key innovation is the AI's ability to orchestrate multiple specialized services through the relationship trust network, creating solutions that no single vendor could provide alone.
MCPI is an open protocol in its early stages. We're looking for contributors to help shape its development!
Contribute on GitHub