Reading up on Model Context Protocol
25 deep · digging since nov 25, 25
- GitHub vs Vercel vs Replit: What Dev Platforms Do When AI Code Is Cheap
As AI code generation becomes cheap, GitHub, Vercel, and Replit are shifting focus to orchestration, production deployment, and verification to capture value in software development.
- What Codex Actually Sends to the Model
The author measured Codex’s HTTP request sizes, showing baseline ~9.4k tokens dominated by built‑in instructions and tools, growing with file reads, command output, images, and history compaction.
- persistent.info: Sierra MCP Gateway Dev Diary
The diary chronicles Sierra engineers’ incremental construction of an MCP gateway, detailing weekly service additions, user adoption, and solutions for secure data access.
- Introducing the Safari MCP server for web developers
Apple released a Model Context Protocol server for Safari that lets AI agents inspect DOM, network requests, screenshots, and console output to debug websites autonomously.
- The Shift to Multiplayer Work: Say Hello to Slackbot’s MCP Client
Slack launches an MCP client for Slackbot, enabling a single conversational interface to connect fragmented enterprise tools and shift work from private silos to shared team channels.
- Introducing the MDN MCP server
Mozilla's MDN MCP server gives AI coding agents real-time access to accurate web documentation and browser compatibility data, outperforming static model knowledge.
- AgentGrade | AI agent readiness scanner
AgentGrade scans websites for AI agent readiness — MCP, llms.txt, payment protocols — and provides a graded report with remediation guides.
- The evolution of agentic surfaces: building with Claude Managed Agents
Claude Managed Agents decouples model reasoning from code execution to let teams deploy production-grade agents without building custom infrastructure.
- GitHub - Ataraxy-Labs/sem: Semantic version control => entity-level diffs, blame, and impact analysis on top of git. 26 languages via tree-sitter. Built for coding agents.
sem is an open-source CLI that replaces line-level git diffs with entity-level awareness (functions, classes, methods) using tree-sitter parsing, designed for both human developers and AI coding agents.
- AI: Morgan Stanley to open its wealth management funnel to agents
Morgan Stanley will let external AI agents pull client stock-plan data directly from ShareWorks and Equity Edge, bypassing human interfaces.
- How we contain Claude across products
Anthropic details how it contains AI agent blast radius across three products using sandboxes, VMs, and human-in-the-loop controls, while documenting missed risks like pre-trust code execution and user-mediated prompt injection.
- Show HN: Apfel – The free AI already on your Mac
Apfel is a CLI tool that gives zero-config access to the on-device 3B-parameter LLM built into macOS Tahoe, requiring no downloads or API keys.
- How Stripe’s Minions Ship 1,300 PRs a Week
Stripe's unattended 'Minion' coding agents merge over 1,300 PRs weekly by leveraging pre-existing developer infrastructure, hybrid blueprints, and layered feedback loops.
- @chrlschn - MCP is Dead; Long Live MCP!
CLIs offer token savings for known tools, but MCP over streamable HTTP is essential for organizational agentic engineering due to centralization, auth, telemetry, and dynamic content delivery.
- The era of “AI as text” is over. Execution is the new interface. - The GitHub Blog
GitHub's Copilot SDK lets developers embed agentic planning-and-execution loops into applications, shifting AI from text responses to programmable infrastructure.
- MCP is dead. Long live the CLI
The Model Context Protocol (MCP) offers no real benefit over command-line interfaces (CLIs), which are more composable, debuggable, and reliable for LLM tool use.
- Anthropic integrates interactive MCP apps into Claude
Anthropic updated Claude to let users interact with Asana, Slack, Figma, and Box tools directly in chat using interactive MCP apps, enabling real-time collaboration on content.
- CLAUDE_AGENT_SDK_SPEC.md
The Claude Agent SDK specification defines a JSON-message protocol for programmatic access to Claude Code's agentic capabilities via stdin/stdout.
- Introducing MCP CLI: A way to call MCP Servers Efficiently
mcp-cli reduces MCP-related token usage by 99% through dynamic discovery, letting AI agents load only needed tool definitions instead of all upfront.
- Cursor agent best practices
Cursor's official guide details best practices for coding agents, emphasizing planning, context management, rules, and iterative review to maximize productivity.
- Claude Code and What Comes Next - by Ethan Mollick
Claude Code, using agentic harness and compaction, autonomously built a working website and game, showing a major leap in AI capability for sustained work.
- Announcing official MCP support for Google services
Google launches fully-managed remote MCP servers for its cloud services, enabling AI agents to interact with tools like Maps, BigQuery, and GKE through a unified, governed endpoint.
- GitHub - addyosmani/gemini-cli-tips: Gemini CLI Tips and Tricks
Gemini CLI offers roughly 30 pro-tips for agentic coding, covering persistent context, custom commands, MCP extensibility, memory, and checkpointing for safe development workflows.
- The Bitter Lesson of LLM Extensions
The evolution of LLM extensions shows that giving agents general-purpose computing abilities is more effective than specialized tool protocols like MCP.