Your Agent Is an Infinite Canvas — RL Nabors, Dressed for Space

2026-05-23 · Source: AI Engineer · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

Rachel Lee Neighbors discusses the evolving landscape of web browsing and content consumption through AI agents, drawing on her experience with web standards and AI startups. She focuses on the Multi-Client Protocol (MCP) as a means for agents to interact with web content. The presentation contrasts STDIO and HTTP transports for MCP servers, advocating for HTTP due to its user experience benefits and compatibility with serverless architectures. Neighbors introduces MCP tools, which return structured data like JSON or markdown, and interactive MCP apps, single-file bundles of HTML, CSS, and JavaScript that provide rich media experiences within agent interfaces. She also details Web MCP, a browser-native approach that allows HTML pages to expose functions as agent tools, available in declarative and imperative models. The discussion underscores the potential of existing browser APIs for creating dynamic agentic experiences, despite current challenges in MCP specification and client implementation.

Key takeaway

For AI Engineers and Software Engineers developing agentic applications, you should recognize that the web's existing infrastructure and APIs provide a robust "infinite canvas" for interactive agent experiences. Explore Multi-Client Protocol (MCP) and Web MCP to integrate dynamic web content, including rich media and interactive apps, directly into agents. This approach allows you to move beyond basic text-based interactions, leveraging browser primitives like Web Speech and Canvas, and ensures your agentic solutions are more engaging and functional. Ignoring these capabilities risks limiting your agents to a suboptimal, text-only user experience.

Key insights

The web is an infinite canvas, and existing browser primitives can transform agents into interactive content platforms.

Principles

• Meet your audience where they are, including in agentic environments.
• Design systems streamline MCP app development and consistency.
• MCP apps are sandboxed; manage external resources via CSP.

Method

Host MCP servers via HTTP endpoints for serverless compatibility. Define MCP tools with input schemas, returning structured data or interactive MCP apps by pointing a "meta" attribute to a UI URL. For browser-native agent tools, use Web MCP's imperative navigator.modelContext.registerTools.

In practice

• Prefer HTTP transports for MCP tools to enhance user experience in agent configurations.
• Bundle MCP apps as single HTML files, embedding resources or managing external links via CSP.
• Build MCP apps using a design system to share fonts, CSS, and components efficiently.

Topics

• Multi-Client Protocol
• Web MCP
• AI Agents
• Browser APIs
• Agentic Interfaces
• Design Systems

Best for: AI Architect, AI Product Manager, AI Engineer, Software Engineer

Related on AIssential

• The Agent Client Protocol Overview — ACP standardizes AI agent-client communication, enabling seamless integration across diverse coding assistants and IDEs.
• MCP Marketplace Brings Real-Time Intelligence to Agentic Applications — Agentic applications require real-time external intelligence to move beyond historical data and make informed, autonomous decisions.
• The Future of MCP — David Soria Parra, Anthropic — MCP provides a universal protocol for agents to ship interfaces and tools, enabling broad connectivity and rich semantic interactions.
• MCP in Claude Code — MCP is an open protocol enabling agents like Claude Code to access and utilize external tools via dedicated servers.
• MCP Dev Summit [Day 2] ft AWS, Docker, & Datadog — MCP is evolving to support real-time, interactive, and secure agent experiences, moving beyond simple request-response models.
• Building agents that reach production systems with MCP — MCP standardizes agent-to-system integration, enabling scalable, portable, and context-efficient connections for cloud-based AI agents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Engineer.