The moment you arrived · Gauri

HomeWorkBlogResumeContact

GauriGauri

© 2026 GauriAhmedabad, GujaratUpdated November 2026
WorkWritingResumeContact
in𝕏
All postsmodelcontextprotocol.io

AI tooling

MCP, explained: how AI tools plug into your stack

The Model Context Protocol is the reason one AI agent can talk to your database, your docs and your dev server. Here is what it is and how to use it safely.

Nov 2026·5 min read
Illustration of a chip-filled brain inside a light bulb on a blue and purple gradient
Photo: Omar Lopez-Rincon on Unsplash

The problem it solves

Every AI coding tool wants the same things: read your files, query your database, look at your running app, open an issue. Before MCP, each tool built its own one-off integration for each service, so connecting N tools to M services meant N × M pieces of glue.

The Model Context Protocol fixes that with one shared interface. The official site describes it as "an open-source standard for connecting AI applications to external systems." Anthropic introduced it in late 2024, and it has since become the common plug that most agents speak.

How it works

An MCP server wraps a system (a database, a design tool, your framework's dev server) and exposes what it can do as tools and resources. An MCP client, built into the AI app, discovers those capabilities and calls them when the model decides they help. Write the server once and every MCP-capable tool can use it.

  • Claude Code's docs describe MCP as "an open standard for connecting AI tools to external data sources" and use it for integrations.
  • Google's Antigravity app lists MCPs alongside Skills and hooks as ways to customise its agents.
  • Next.js 16 shipped a Next.js DevTools MCP, so an agent can inspect your running Next.js app while it debugs.

Adding a server to a project

In Claude Code, project-level servers live in a .mcp.json file at the repo root, so the whole team shares the same setup. A minimal example that wires up the Next.js DevTools server:

json
{
  "mcpServers": {
    "next-devtools": {
      "command": "npx",
      "args": ["-y", "next-devtools-mcp@latest"]
    }
  }
}

Use it safely

An MCP server is code that runs with your permissions and feeds text straight to a model, so treat it like any other dependency.

  • Install servers only from sources you trust, and pin versions for anything that touches production data.
  • Give each server the narrowest access it needs: a read-only database user, a scoped API token.
  • Remember that tool output is untrusted input. A web page or an issue body can contain instructions aimed at the model, so keep a human in the loop for anything destructive.

Sources & further reading

  • Model Context Protocol: official site
  • Claude Code overview (Anthropic docs)
  • Antigravity app (Google)
  • Next.js 16 release post
Older AI coding agents in late 2026: from one assistant to a team
Back to all postsDiscuss on email ↗