How it works
Anthropic introduced MCP in November 2024 and in December 2025 donated it to the Agentic AI Foundation, a Linux Foundation fund that Anthropic co-founded with OpenAI and Block. An MCP server exposes tools (actions the model can call), resources (data it can read) and prompts (reusable templates). An MCP client inside a host app, such as a chat app, an editor or an agent, discovers them and offers them to the model, exchanging JSON-RPC messages.
Local servers run as a process on your machine and talk over standard input and output; remote servers are reached over HTTP (Streamable HTTP) and usually sign users in with OAuth. Thousands of servers exist, for GitHub, databases, file systems, browsers, Slack and more. A server can do anything its credentials allow, and its tool descriptions go straight into the model's context, so install only servers you trust and give them the narrowest access that works.
Model Context Protocol (MCP) pros and cons
Pros
- Open, vendor-neutral standard backed by the major AI companies
- Write a tool integration once and use it from many apps
- Large and growing catalogue of ready-made servers
- Works locally for private data or remotely as a hosted service
Cons
- A malicious or careless server can leak data or take harmful actions
- Many tools in context use up tokens and can confuse the model
- The specification still changes, and older servers lag behind it
When to use Model Context Protocol (MCP)
Pick it when
- You want your product's data or actions usable from AI assistants
- An internal agent needs the same tools across several apps
- Connecting a coding assistant to docs, databases or issue trackers
Skip it when
- One app calls a few tools of its own (plain tool calling is simpler)
Model Context Protocol (MCP) pricing
Open source
Free: the specification and official SDKs are open source.
Approximate, checked September 2026.What the other tools cost
Related terms
More in AI and LLMs
Building with models