Citation :

Tanishka Dhanwan , 2026. "Designing Web Interfaces for Governed LLM Agents: A Review of MCP, Structured Access, and Human Control" ESP International Journal of Advancements in Science & Technology (ESP-IJAST)  Volume 4, Issue 4: 25-32.

Abstract :

Websites that can be used by language-model agents need more than a collection of API endpoints. They need compact observations, unambiguous actions, current state, authorization boundaries, accessible human fallbacks, and evidence that can be checked after execution. This review examines ten supplied studies covering agentic AI, Model Context Protocol (MCP), structured web access, human–AI dialogue, tool security, and domain-specific orchestration. It separates findings that recur across the literature from claims that remain dependent on a particular benchmark or implementation. The review then organizes the evidence into a layered design in which an LLM interprets and plans, while policy gateways and deterministic services validate, authorize, execute, and verify. It concludes with a research agenda for effectiveness, context cost, latency, security, accessibility, and maintainability. The paper does not claim that MCP alone improves every website.

References :

[1] Sarfraz Brohi and Qurat-ul-ain Mastoi and N. Z. Jhanjhi and Thulasyammal Ramiah Pillai. “A Research Landscape of Agentic AI and Large Language Models: Applications, Challenges and Future Directions.” Algorithms, 2025.

[2] Suhani Mohanty and Rayner Dcunha and Marek Z. Reformat and Petr Musilek. “ARGUS: An MCP-Based Agentic LLM Architecture for Distribution Grid Decision Support.” 2026.

[3] Luca Marconi and Luca Longo and Federico Cabitza. “Assessing Interaction Quality in Human--AI Dialogue: An Integrative Review and Multi-Layer Framework for Conversational Agents.” Machine Learning and Knowledge Extraction, 2026.

[4] Rayhan Zahwan Saleh and Muharman Lubis. “Design and Implementation of MCP-Web-Curl: A Model Context Protocol Server for Web and API Access in Agentic Coding Assistants.” Jurnal Teknik Informatika, 2026.

[5] Andrei Kojukhov and Ilya Levin and Arkady Bovshover. “From Subsumption to Semantic Mediation: A Generative Orchestration Architecture for Autonomous Systems.” Algorithms, 2025.

[6] Carlos Avila and Daniel Ilbay and David Rivera. “Human--AI Teaming in Structural Analysis: A Model Context Protocol Approach for Explainable and Accurate Generative AI.” 2025.

[7] Jingfeng Zhou and Jiying Chen and Meng Wang and Baihua Dong and Xiaochi Zhou. “Large Language Model Orchestrated Workflow Integrating Retrieval-Augmented Generation and Model Context Protocol Tools for Urban-Scale Simulation-Based Resilience Assessment.” Building and Environment, 2026.

[8] Aaron Steiner and Ralph Peeters and Christian Bizer. “MCP vs RAG vs NLWeb vs HTML: A Comparison of the Effectiveness and Efficiency of Different Agent Interfaces to the Web.” 2026. DOI: 10.1145/3774904.3792893.

[9] Mohammed Mehedi Hasan and Hao Li and Emad Fallahzadeh and Gopi Krishnan Rajbahadur and Bram Adams and Ahmed E. Hassan. “Model Context Protocol (MCP) at First Glance: Studying the Security and Maintainability of MCP Servers.” 2026.

[10] D. Perera. “WEBMCP: Efficient AI-Native Client-Side Interaction for Agent-Ready Web Design.” 2026.

[11] Model Context Protocol. “Model Context Protocol Specification.” 2025. Available: https://modelcontextprotocol.io/specification/2025-06-18.

[12] Model Context Protocol. “Architecture Overview.” 2026. Available: https://modelcontextprotocol.io/docs/2026-07-28/learn/architecture.

[13] [13] {Model Context Protocol. “Security Best Practices.” 2026. Available: https://modelcontextprotocol.io/docs/2026-07-28/tutorials/security/security_best_practices.

[14] {OWASP Foundation. “AI Agent Security Cheat Sheet.” 2026. Available: https://cheatsheetseries.owasp.org/cheatsheets/AI_Agent_Security_Cheat_Sheet.html.

[15] {World Wide Web Consortium. “Web Content Accessibility Guidelines (WCAG) 2.2.” 2024. Available: https://www.w3.org/TR/WCAG22/.

[16] {Google Chrome Developers. “How the Core Web Vitals Metrics Thresholds Were Defined.” 2025. Available: https://web.dev/articles/defining-core-web-vitals-thresholds.

[17] {Schema.org Community. “Schema.org.” 2026. Available: https://schema.org/.

[18] Thibault Le Sellier De Chezelles and Maxime Gasse and Alexandre Drouin and others. “The BrowserGym Ecosystem for Web Agent Research.” arXiv preprint arXiv:2412.05467, 2024. Available: https://arxiv.org/abs/2412.05467.

Keywords :

Model Context Protocol, Web Agents, Structured Interfaces, Agent Security, Accessibility, Human–AI Interaction.