The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in November 2024, that lets AI applications connect to external tools, data sources, and systems through a uniform interface.
MCP standardizes how AI agents access tools and data. An AI application (the host) runs MCP clients that connect to MCP servers, each exposing a tool or data source over JSON-RPC 2.0. This replaces custom one-off integrations with a single protocol. Because MCP servers can take real actions, authenticating and authorizing those connections is a core security concern.
Key Takeaways
- MCP is an open standard from Anthropic (November 2024) for connecting AI models to tools and data.
- It uses a host, client, and server model communicating over JSON-RPC 2.0.
- MCP solves the “N times M” problem of building a custom connector for every tool.
- It has been widely adopted, including by other major AI providers, as a de facto standard.
- MCP servers can take real actions, so authentication, authorization, and auditing are essential.
What Is the Model Context Protocol (MCP)?
MCP is an open standard that defines a uniform way for AI applications to connect to external tools, data sources, and systems. Instead of each AI product building bespoke integrations, MCP gives them a common protocol, so an AI agent can discover and use any tool that exposes an MCP interface. Anthropic introduced it in November 2024, and it has since been adopted across the industry.
Why MCP Exists
Before MCP, connecting AI models to tools was an N times M problem: every model needed a custom connector for every tool or data source, which did not scale. MCP collapses that into a single standard. A tool builder writes one MCP server, and any MCP-compatible agent can use it. This is the same kind of simplification that standard protocols brought to other parts of computing.
How MCP Works
MCP uses a host, client, and server model:
- Host: The AI application (for example, an assistant or agent) that needs external capabilities.
- Client: A connector the host runs, one per server, that manages the connection.
- Server: A component that exposes a tool, data source, or other capability.
Clients and servers exchange JSON-RPC 2.0 messages over one of two standard transports: stdio for local servers and Streamable HTTP for remote ones (the older HTTP+SSE transport is now deprecated). The host creates a dedicated client for each server it connects to, keeping connections isolated.
What MCP Servers Expose
MCP servers can offer three kinds of capability: tools (actions the agent can invoke, such as querying a database or sending a message), resources (data the agent can read), and prompts (reusable templates). Because tools can change real systems, the actions an MCP server allows define how much trust it requires.
Securing MCP
Since MCP servers can take real actions and reach sensitive data, they must be secured like any privileged machine-to-machine connection. The protocol now supports this directly: the November 2025 specification added server identity verification and OAuth-based machine-to-machine authorization, including client-credentials and enterprise identity-provider controls. Putting that into practice means giving each server a verifiable identity, scoping its permissions to least privilege, authenticating every connection, and auditing what agents do through it. This is the machine-identity discipline described in non-human identity, applied to MCP.
MCP and agentic AI
MCP is a foundational piece of agentic systems because it is how agents reach tools and data. Securing MCP connections is therefore part of agentic AI security, and the two topics are best planned together.
How Encryption Consulting helps
As AI agents and MCP servers multiply, each becomes a machine identity that must be authenticated and governed. CertSecure Manager and Encryption Consulting’s PKI Services provide the certificates and issuing infrastructure to give MCP servers and agents verifiable, manageable identities, backed by ISO/IEC 27001:2022 and SOC 2 certified practices.
Frequently Asked Questions
Who Created MCP?
The Model Context Protocol was created by Anthropic, which introduced it as an open standard in November 2024. It was developed by Anthropic engineers David Soria Parra and Justin Spahr-Summers. Since release it has been adopted broadly across the industry, including by other major AI providers. In December 2025, Anthropic donated MCP to the Agentic AI Foundation, a fund under the Linux Foundation co-founded with Block and OpenAI, moving it from an Anthropic-maintained project to vendor-neutral governance. It remains an open specification with SDKs for major programming languages.
What Problem Does MCP Solve?
MCP solves the integration problem of connecting AI models to the many tools and data sources they need. Before MCP, each combination of model and tool required a custom connector, an “N times M” problem. MCP replaces those one-off integrations with a single, uniform protocol, so any MCP-compatible agent can talk to any MCP server.
Is MCP Secure?
MCP itself is a protocol; security depends on how MCP servers are deployed and controlled. Because servers can take real actions and access data, they should have verifiable identities, least-privilege permissions, authenticated connections, and full audit logging. Treating MCP servers as privileged machine identities, rather than open endpoints, is essential to using MCP safely.
What Is an MCP Server?
An MCP server is a component that exposes a specific capability, such as a tool, an API, or a data source, to AI applications through the Model Context Protocol. An AI host runs MCP clients that connect to these servers over JSON-RPC 2.0. As of late 2025, more than 10,000 active public MCP servers exist, ranging from file systems and databases to SaaS integrations.
How Does MCP Relate to AI Agents?
MCP is how many AI agents reach the outside world. An agent uses MCP clients to connect to MCP servers that provide tools and data, letting the agent take actions beyond text generation. This makes MCP central to agentic systems, and it makes securing MCP connections a core part of agentic AI security.
Secure Your MCP Connections
Ready to give MCP servers and agents managed identities? See CertSecure Manager in action, or read about agentic AI security.