MCP
29 articles tagged with this topic
WPS Multidimensional Table Runs Python: Kingsoft Quietly Pivots to Platform
WPS Multidimensional Table adds Python, MCP, and 70+ APIs. Kingsoft pivots to a developer platform, but near-zero AI buzz leaves its ecosystem prospec
LangChain: AI Agents Load Skills On-Demand — Modular Dev Is the New Agent Paradigm
LangChain DeepAgent: AI agents load skill modules on-demand like humans, shifting Agent development from monolithic to pluggable composition for custo
DeepSeek-TUI Tops GitHub at 2434 Stars: Terminal AI Agent Goes Practical
Terminal AI agent DeepSeek-TUI installs in 15s, supports MCP & sub-agents. Terminal AI crosses from "works" to "works well"; Chinese LLMs now compete
Google Cloud + 5 Security Firms Build Agent Firewall — AI Stuck on Security Not Tech
Google Cloud + 5 security vendors for Agent Gateway, tackling AI Agent data leaks and tool abuse—enterprise AI bottleneck shifts from tech to trust.
OpenCode Skills: AI Assistants Split 'Knowing' From 'Doing' On Demand
OpenCode Skills inject AI knowledge via Markdown on demand, splitting "manual" from "toolbox." It solves context crowding but adds maintenance overhea
opencode Adopts MCP Protocol: AI's "USB Port" for External Tools Starts to Unify
opencode updates MCP config guide for local and remote servers. The protocol cuts AI-tool integration costs, but data security risks and big-tech ecos
Archon Goes Viral: Ditch AI Free-Play, Deterministic Orchestration Is Endgame
Archon drops AI free-play for deterministic workflows. This "code does dirty work, AI thinks" hybrid is the sole fix for enterprise AI black-box chaos
GitNexus Gives AI Coders the Big Picture — Open Source Tackles Blind Code Edits
GitNexus indexes repos into knowledge graphs via MCP, letting AI check impact before editing. Zero token indexing—a leap from writing to reading code.
200 Lines of Code to Let AI Control Your PC—Agent Deployment Still Stuck on Security
Indie developer uses Node.js to let DeepSeek operate a Linux VM. This minimal experiment reveals Agent deployment's core contradiction: easy capabilit
Semvec Ends AI Chat Cost Explosion — Long-Context Memory Becomes New Track
Semvec swaps chat history for fixed semantic states, cutting tokens 76% over 48 rounds. AI savings shift from cheap models to smarter memory.
Open-Source Hybrid Recall Tool Gives Agents Memory Without Giant Contexts
Qwen3.5-4B MCP tool uses BM25+vector hybrid recall for Agent project memory. Focus shifts from "bigger context" to "better retrieval," cutting deploym
Warp Open-Sources AI Terminal: The 40-Year-Old Black Box is Finally Rebuilt
AI terminal Warp open-sourced its Rust client. It turns the 40-year-old CLI into blocks for direct AI debugging and deployment. Backed by Sequoia and
Deconstructing the LLM Lineage: From LLM to Agent, It's All Context Patching
From RAG to MCP, buzzwords overwhelm. We map the core logic: LLMs just predict text; later tech patches their gaps. Grasp this, and jargon won't fool
LangChain Standardizes AI Tool Calling: LLMs Shift from Talking to Doing
LangChain updates tool APIs for LLMs to interact with external systems. AI shifts from chatbots to executors; tool calling is key to enterprise AI ado
Anthropic's App Store for AI Coding: Skills Shift from Code to Workflows
Claude Code's plugin marketplace uses Agent Skills to shift AI from code to workflows. Ecosystems grow, but third-party quality and security remain un
Yank Note Adds MCP: Local Docs Now Act as AI's Hands and Feet
Yank Note adds MCP and sidebar Agent, making local notes a callable AI entry point. Knowledge tools are evolving from display containers into AI execu
Anthropic Unpacks Claude Code: AI Coding Shifts from Chat to Building Blocks
Anthropic unpacks Claude Code's command architecture, adding Skills and MCP. AI coding shifts from chat to permission-controlled modular workflows, se
Microsoft Red Teams 100 AI Agents: Single Safety ≠ Network Security
Microsoft's 100-agent red team test found four network-only risks: worm spread, info amplification, trust hijacking, and stealth. Single safety ≠ netw
Anthropic Adds Four-Layer Memory to Claude - AI Coding Tools Work With Context
Claude Code unveils four-layer memory architecture auto-injecting preferences and project rules per conversation, with MCP IDE integration.
Introducing the Agent Readiness score. Is your site agent-ready?
Cloudflare's isitagentready.com scans sites for AI agent compatibility; only 4% of top 200K domains declare AI preferences .
Nous Research Open-Sources Her mes Agent, a Self-Improving AI Agent Framework
Hermes Agent hits 90 K+ GitHub stars with persistent skill memory and three-layer architecture across 200+ models.
Claude Code Skills vs MCP: Architecture Deep Dive
Anthropic's Claude Code uses two distinct extension layers: Skills for reusable domain workflows and MCP for real-world tool connectivity.
Claude Code MCP Plugin Architecture: Cross -Process Tool Proxy Dissected
Source analysis reveals Claude Code uses stdio- based MCP servers as isolated subprocesses, proxying external tools transparently to the model.
everything-claude-code: 181-Skill Claude Code Enhancement Framework
Open-source framework adds 181 skills , 47 sub-agents, and 34 rules on top of Claude Code CLI.
ByteDance Open-Sources DeerFlow 2.0: 18 -Layer Agent Middleware Runtime
ByteDance's DeerFlow 2.0 hit GitHub Trending #1 on Feb 28, 2026, rebranding from a research framework to a full agent runtime infrastructure with an
Anthropic Decouples Agent Brain From Execution in Managed Agents Arch
Anthropic's new architecture splits Claude agent systems into five isolated components to enable fault recovery and credential isolation.
MCP Protocol Standardizes AI-to-External-System Connections
Model Context Protocol gives AI models a USB-like standard interface to external APIs, databases, and tools.
Amazon Bedrock AgentCore Gateway Now Supports OAuth 2.0 for MCP Servers
AgentCore Gateway centralizes MCP server auth using OAuth 2.0 Authorization Code flow, removing per-server credential management.
Claude Code Queries 15 Repos to Deliver Hyper-Personalized Support
A non-engineer at Galileo built a Claude Code system querying 15 repos plus Confluence to answer enterprise customer questions without engineering esc