What is an MCP server?

What MCP means

MCP, or Model Context Protocol, is an open-source standard for connecting AI applications to external systems (MCP introduction). It gives applications a shared way to connect to data sources, tools, and workflows (MCP introduction).

Host, client, and server

The host is the AI application. It creates and manages MCP clients, and creates one client for each connected server (MCP architecture).

An MCP client maintains a connection to a server and obtains context for its host. The server is the program that provides that context (MCP architecture).

What an MCP server can provide

Servers can expose tools, resources, and prompts. Tools perform actions, resources provide contextual data, and prompts are reusable interaction templates (MCP architecture).

For example, MCP can connect AI applications to data sources and tools, helping them access information and perform tasks (MCP introduction). MCP defines how context is exchanged, but does not dictate how an AI application uses its language model or manages that context (MCP architecture).

Local stdio servers

A local server runs on the user's machine. With stdio transport, local processes communicate through standard input and output, without network overhead (MCP architecture). The terms “local” and “remote” describe where the server runs (MCP architecture).

A local server runs with the user's account permissions. Limit its access to files and directories you are comfortable allowing it to read or modify (local server guide).

Remote servers over HTTP

A remote server runs on the internet and can expose tools, prompts, and resources to AI applications (remote server guide). Streamable HTTP uses HTTP POST, with optional Server-Sent Events for streaming; it supports standard authentication methods (MCP architecture).

Remote servers can be available to MCP clients with an internet connection without installation on each device. Most require authentication, such as OAuth or API keys (remote server guide).

Where KnowMan fits

KnowMan's MCP server runs locally with npx -y @knowman/mcp and as a hosted server at https://api.knowman.io/mcp with OAuth sign-in or an API key. It connects to KnowMan's HTTP API and offers tools to search saved items, ask questions with citations, read an item, save a URL, view recent items, and access saved memories. See the KnowMan MCP knowledge base guide.

KnowMan is one option when you want an AI client to work with your saved articles, links, and memories. Its server is local, so it does not fit clients that accept only remote MCP servers. For more on connecting AI tools to notes, see connect AI to your notes.

Connect to the hosted KnowMan server

Knowing the difference between stdio and HTTP transports, you can try both with KnowMan. For the remote route, add https://api.knowman.io/mcp to a client that accepts hosted MCP servers. On first use, the client opens the KnowMan consent page, where you sign in and approve the requested scopes: knowman.read for search, read, and ask, and knowman.write for saving links. Access tokens last one hour and refresh automatically.

You can also send an API key instead of signing in with OAuth. Create the key in KnowMan settings and pass it as an Authorization: Bearer <key> header, for example with Claude Code: claude mcp add --transport http knowman https://api.knowman.io/mcp --header "Authorization: Bearer <key>".

For the local route, the open-source client at https://github.com/Essatt/knowman-mcp starts with npx -y @knowman/mcp on Node 22+ and speaks stdio, exposing the same six tools: kb_search, kb_ask, kb_get_item, kb_save, kb_recent, and kb_memories.