Learn how teams at NBIM, Brex, and more build reliable AI agents with Claude on AWS Bedrock.
Learn how teams at NBIM, Brex, and more build reliable AI agents with Claude on AWS Bedrock.
Learn how teams at NBIM, Brex, and more build reliable AI agents with Claude on AWS Bedrock.
Connect AI assistants to your tools without custom integrations using Model Context Protocol.
AI models are only as good as the context provided to them. AI assistants like Claude can answer questions and perform an impressive range of tasks, but if they can't access the data or tools they need, they're limited in what they can do for you. You typically solve this by copying and pasting context from one tab to another, whether it's editing a document in Google Drive, replying to a thread in Slack, or updating code in an IDE. This process is slow, manual, and risks leaving out important context.
The Model Context Protocol (MCP) offers a solution that is open and widely available across all AI apps and assistants. In this article, you'll learn what MCP is, how it works and why it matters, and who it's for. You'll see examples of MCP in action and understand how you can start using or building with MCP today.
The Model Context Protocol is an open standard that defines how LLMs communicate with external systems.
Think of MCP as USB-C for LLMs. Just as USB-C provides a universal connector for your phone, laptop, and other devices, MCP provides a universal format for LLMs to connect with external systems. Before USB-C, every electronic gadget had its own cable: Lightning for iPhone, micro-USB for Android, proprietary connectors for cameras. As more devices adopted USB-C, connectivity became seamless across the ecosystem.
MCP brings this same simplicity to AI integrations. Before MCP, every application and database required custom code to connect with LLMs. Google Drive needed its own integration, Slack needed another, Figma yet another. Now, MCP provides a single, standardized format for connecting these tools to Claude and other AI applications.
MCP was created at Anthropic by David Sorria Para and Justin Spahr-Summers. The idea originated from David's frustration with constantly copying code between Claude Desktop and his Integrated Development Environment (IDE). Recognizing this as a classic M×N problem where multiple applications need multiple integrations, David pitched building a protocol to solve this to Justin. They designed MCP based on the popular Language Server Protocol and open-sourced it in November 2024 with Anthropic's support to ensure the entire AI ecosystem could benefit.
MCP works through a two-sided approach. AI agents and chatbots like Claude create MCP Clients, so they can connect to applications like Notion, Canva, or Figma, who make their tools and data available through MCP Servers.
By building an MCP Client, AI agents and chatbots can access thousands of MCP Servers built by the community, giving them a straightforward path to extend their capabilities. By building an MCP Server, companies and developers can make their products readily available to AI, creating a new avenue to provide value.
As MCP is open-source, anyone can build an MCP Server or Client.
MCP allows LLMs to go beyond chat and perform real-world tasks: reading an email thread and sending a reply, accessing a codebase and deploying an update, or reviewing a design brief and generating a first draft. The protocol creates a foundation for LLMs to connect with external systems, tools, and applications to access data and take actions. This provides:
AI assistants gain access to thousands of tools — Once an AI assistant implements MCP (via an MCP client), it can instantly connect to thousands of MCP-compatible applications, from specialized coding tools to enterprise workflow platforms, without building custom integrations for each one.
Tools and applications connect to every AI assistant at once — Companies like Notion, Figma, or Asana build a single MCP server that works with any AI assistant that’s compatible (i.e. has implemented an MCP client). Developers only need to build one integration for all AI connections.
Anyone can build and share — As an open standard, MCP servers published by developers or companies are compatible with any MCP client. This openness has created a thriving ecosystem of thousands of community-built servers, accelerating the availability of tools and applications for AI assistants..
Makes software AI-accessible by design — Traditional software is built for humans using web interfaces. MCP provides a parallel interface designed for AI interaction, allowing applications to become truly AI-native. This means better, more reliable integrations between AI models and the tools people already use.
MCP creates the infrastructure for AI agents to access any number of services and tools, creating true end-to-end task automation. As more applications adopt the protocol, the vision of AI agents that can independently handle complex, multi-step workflows becomes increasingly practical.
Developers get a standardized way to build integrations once and have them work with any compatible AI. Enterprises gain secure, IT-controlled AI connectivity that scales across their organization. Consumers can connect their favorite tools to AI instantly, with no technical knowledge required.
Developers can follow a single standard to connect external products to your AI applications and agents. This simplifies the process of building integrations, grows the number of available products to connect to, and improves the overall quality and security of connectivity in the ecosystem.
Building an agent that will connect to many applications? Building an application that will connect to many agents? MCP provides you with access to an ecosystem of compatible tools with streamlined integration.
Enterprises can drive internal adoption of AI tools and applications more effectively, as MCP simplifies the process of connecting your systems to AI. This helps make AI more connected within your organization, expanding its capabilities and usefulness for your staff.
MCP provides end-users with seamless connectivity between their favorite AI assistants and work tools. It makes it easier to automate tasks and avoid copying and pasting across tabs. In short, MCP gives AI greater access and connectivity to your world.
In Claude, you can instantly connect to MCP Servers, known as Connectors. This provides you with a straightforward way to connect Claude to your favorite work apps.
The real value of MCP becomes clear when you see it in action with the tools you already use. Here are some examples of MCP being used to power integrations in Claude, known as Connectors:
The Canva Connector allows Claude to generate new designs directly within Canva. Using MCP, Claude can connect to the tools Canva provides to generate designs on the canvas.
Using the Notion and Linear Connectors, Claude can access your pages in Notion and use them to update tickets in Linear. Here MCP creates a seamless transfer of unstructured context into organized tickets in a separate project management system.
The Figma Connector allows Claude to access designs within Figma. This lets Claude Code create working prototypes of websites, applications, or user interfaces based on designs created in Figma.
Claude Connectors include integrations for:
Each connector takes just a few seconds to configure to become part of Claude's working context. Outside of Claude, there is an ecosystem of MCP servers on the open-source MCP Registry.
Two paths exist based on your needs.
Connectors are pre-built, giving Claude instant access to tools, databases, and applications, and providing you with a new set of capabilities. Open Claude, browse available connectors, and click to add them.
MCP is open-source, meaning that anyone can adopt MCP to connect AI to applications. The Model Context Protocol documentation walks through how to build with MCP.
If you want to try MCP, start by browsing for a Claude Connector you can immediately start using with Claude.
If an existing MCP server doesn't already exist, creating your own takes some work, but isn't too complex if you know TypeScript or Python. The Model Context Protocol quickstart has working examples you can modify for your needs.
Get the developer newsletter
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
Try out MCP Connectors in Claude.ai
Try out MCP Connectors in Claude.ai
Try out MCP Connectors in Claude.ai
Get the developer newsletter
Product updates, how-tos, community spotlights, and more. Delivered monthly to your inbox.
