An MCP server is a program that exposes data and capabilities to AI applications over the Model Context Protocol, advertising them as tools (callable functions), resources (readable data such as files or records), and prompts (reusable templates). AI clients connect to it over JSON-RPC and negotiate which of these capabilities are available.
Why it matters
The MCP server is the unit of integration in the MCP ecosystem: it is what turns a database, API, or document store into something an AI assistant can actually read and act on. Because the interface is standardized, one server can serve many different AI clients, and an organization can wrap its internal systems once rather than per-tool. This makes proprietary or private knowledge usable by AI assistants without exporting it into the model's training data.
How Pith relates
Pith's MCP server exposes your reading memory, the auto-built cited wiki, your bookmarks, and search, so a connected AI assistant can retrieve grounded answers from what you have actually saved. Every result the server returns carries its source, keeping the assistant's responses traceable. You connect the server once and reuse it across any MCP-compatible client.
See also
Last reviewed: 7 June 2026 · Licensed CC BY 4.0 · cite freely with attribution to Pith.