An open standard that lets AI models connect to tools and data sources through one consistent interface.
Official siteThe Model Context Protocol (MCP) is an open standard, introduced by Anthropic in late 2024, for connecting AI models and agents to external tools, data and systems. It plays a role often described as "USB-C for AI": instead of writing a bespoke integration for every model-to-system pairing, you expose capabilities once as an MCP server and any MCP-compatible client can use them.
An MCP server advertises tools (actions the model can call), resources (data it can read) and prompts. This decouples what a model can do from how it is wired up, and it has been adopted across the major AI ecosystems, making it a practical backbone for real agent work.
MCP defines a client-server protocol between an AI application (the client) and capability providers (servers). A server exposes tools, resources and prompts over a standard transport; the model discovers and invokes them without knowing the implementation details behind each one.
When we build AI agents that must act on real systems — query a database, call an internal API, read documents — MCP lets us expose those capabilities once and reuse them across agents and models, instead of hard-wiring each integration.
MCP is young, so tooling and conventions are still settling and the security model (what an agent is allowed to invoke) demands careful design. For a single throwaway integration, a direct API call is simpler. The standard pays off once you have several tools and want them portable across agents.
Our AI/agent work uses TypeScript and Node to build MCP servers that expose internal tools and data to agents in a controlled, reusable way — the same pattern behind our RAG knowledge assistants, where retrieval is exposed as an MCP capability.
The Model Context Protocol (MCP) is an open standard, introduced by Anthropic, for connecting AI models and assistants to external tools, data sources and systems in a consistent way. It matters because AI assistants are only as useful as the context and capabilities they can access — an assistant that can read your documents, query your database or call your APIs is far more valuable than one working from its training data alone — and before MCP, every such integration was bespoke. MCP aims to be a universal connector (often likened to "a USB-C port for AI"), letting any compliant model talk to any compliant tool or data source through one standard interface.
MCP defines a client-server architecture: an MCP server exposes a specific capability — access to a set of files, a database, an API, a tool — through the standard protocol, and an MCP client (built into an AI application or assistant) can discover and use any MCP server. This decoupling means a tool built as an MCP server works with any MCP-compatible AI application without custom integration, and an AI application gains new capabilities simply by connecting to more servers. It standardises how context and tools are described, requested and returned, so the messy work of wiring models to the outside world becomes plug-and-play.
MCP matters strategically because it addresses the integration bottleneck holding back practical AI: rather than every AI product building custom connectors to every tool and data source, a shared protocol lets an ecosystem of interoperable servers and clients emerge, much as HTTP and USB unlocked their domains. For teams building AI agents and assistants, it means faster, more maintainable integrations and access to a growing library of ready-made connectors; for tool and data providers, exposing an MCP server makes their system usable by any compliant AI application. As an open standard gaining adoption, it is becoming part of the foundational plumbing of the AI-agent ecosystem.
Part of our technologies knowledge graph — browse every entry in this branch.
A typed superset of JavaScript that catches whole classes of bugs before your code ever runs.
A server-side JavaScript runtime built on V8 that lets one language run your whole stack.
A query language for APIs that lets clients ask for exactly the data they need — no more, no less.
A document database that stores flexible, JSON-like records — great when your data is nested and evolving.
The React meta-framework that adds routing, server rendering and full-stack features on top of React.
Common questions
Straight answers on how this fits your marketing and build.
Still have questions? Talk to a specialist