What is MCP, the Model Context Protocol?

1.What is MCP?

The Model Context Protocol is an open protocol for integrating LLM applications with external data sources and tools; the specification describes it as 'a standardized way to connect LLMs with the context they need.' It takes inspiration from the Language Server Protocol, which standardised how editors support programming languages. The authoritative requirements are defined in a TypeScript schema; the latest revision in the specification repository is dated 2026-07-28.

2.How does it work?

• Hosts are the LLM applications that start connections (an assistant or IDE).
• Clients are connectors inside the host, one per server.
• Servers provide context and capabilities: resources, prompts and tools.
• Clients can offer elicitation, where a server asks the user for more information.

Messages use JSON-RPC 2.0. Optional extensions add features such as Tasks for long-running operations and MCP Apps for interactive UI in a conversation.

{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}

3.What does MCP require from the people who deploy it?

The specification puts consent at the centre: users must explicitly consent to data access and operations, hosts must obtain consent before invoking any tool, and tool descriptions should be treated as untrusted unless they come from a trusted server. MCP itself cannot enforce these principles; implementers are expected to build consent and authorization flows.

4.What does MCP not do?

• It does not make your website readable. Pages still need server-rendered content, llms.txt or markdown.
• It does not run in the page. Tools a web page exposes to an in-browser agent are WebMCP's job. See MCP and WebMCP.
• It does not connect agents to other agents as peers. That is A2A.

5.How do you check an MCP server?

1. Connect with an MCP client and call tools/list. Confirm names and descriptions say what each tool does and when to use it.
2. Call each tool with realistic inputs and confirm errors are readable.
3. Confirm authorization for any tool that touches user data.
4. ora.ai publishes an MCP server of its own that scans a domain and can rescan after each fix; its readiness scan also looks for MCP endpoints.

6.Where does it sit in the readiness stack?

Layer 3, Act: MCP is how most agents call your product's functions.

Frequently asked questions