Today
AnthropicResearchers find Claude Opus four point seven exhibits play-like behaviorSimon Willison's Weblogllm-tools-datasette 0.1.1llama.cpp releasesllama dot cpp build eleven seven two seven fixes silent data corruptionSimon Willison's Weblogllm-echo 0.5MetaOnlyFans creators question Meta allowing Muse AI agent to engage in sextingNvidiavLLM zero point three two ships structured output and broader low-precision KVAnthropicAnthropic grants outside AI watchdog employee-like access to systemsVercelVercel AI Gateway logs now show session affinity and cache hit ratesIEEE SpectrumEdge AI for Smart Cities: From Insight to Cognitive ActionWIREDDisney structured Character dot AI bankruptcy to acquire IP, dodge lawsuitsAnthropicAnthropic commits five percent of shares to new public benefit fundThe VergeWoman arrested in Philadelphia for blocking Uber delivery robot
dailybrief

Agents, MCP & Skills · 2 min read · Updated 2026-10-09

How MCP works: hosts, clients and servers

The three parts

The host is the app, such as a chat assistant or code editor. Inside it, one client keeps a connection to one server. The server is a small program that exposes a tool or a source of data.

How they talk

Messages use JSON-RPC 2.0. When a client connects, it asks the server what it offers. The model then decides when to use something, and the host sends the request.

Local and remote

A local server runs on your own computer and talks over standard input and output (stdio). A remote server runs on the internet and talks over HTTP, often with sign-in. Local is simple; remote is easier to share.

Key takeaways

  • Host runs clients; each client connects to one server.
  • Messages are JSON-RPC 2.0.
  • Servers can run on your machine or online.

Quick questions

Do I have to write a server?

Often not. Many ready-made servers exist. Write one only for your own tool or data.

What is stdio?

Standard input and output: the host starts the server program and swaps messages through it.

Does the model call the server directly?

No. The model asks, and the host app makes the call and returns the result.

Keep learning

What is MCP (Model Context Protocol)?MCP, the Model Context Protocol, is an open standard that lets AI apps connect to tools and data, such as files, databases…2 min readRead →MCP tools, resources and promptsAn MCP server can offer three kinds of things: tools the AI can run, resources it can read, and prompts that are ready-made…2 min readRead →MCP safety: staying in controlConnecting an AI to real tools is powerful and risky. Only use servers you trust, give them the least access they need, and…2 min readRead →