What is MCP (Model Context Protocol)?
MCP is the standard way an AI assistant connects to a tool: a fixed list of things it can do, and a sign-in that says who it is doing them as.
Last checked
MCP, the Model Context Protocol, is an open standard for connecting an AI assistant to a piece of software. The software runs an MCP server, which publishes a short list of tools — things the assistant may ask it to do, each with a name and a description — and the assistant calls them. It is the same idea as a plugin, except that every assistant that speaks MCP can use every server that speaks it, so a tool is built once rather than once per assistant.
The three parts
- A server: the tool’s side. fenbs’s is at
https://fenbs.ai/api/mcp. It lists tools such asfenbs_list_itemsandfenbs_create_item. - A client: the assistant’s side. Claude, Claude Code, ChatGPT, Cursor and VS Code with Copilot all have one built in.
- A sign-in. A remote server needs to know who the assistant is acting as. The good way is OAuth: the client opens the tool in a browser, the person signs in and approves, and the client is handed a token. The bad way is pasting an API key into a config file.
What it looks like in practice
You add one URL to your assistant. A browser opens, you sign in, you tick what the assistant may do, and from then on you can say “what is on my board?” or “file that as a bug” and the assistant calls the right tool. You never see the calls. Every call arrives at the tool with your identity attached, so the tool can apply your permissions and record your name.
How fenbs uses it
MCP is fenbs’s only integration surface, on purpose: one door that every assistant can use, with OAuth on it. An assistant connected over MCP is a member of the board with a role and a name in the history, which is the whole product. Connect steps for Claude, Claude Code, ChatGPT, Cursor and Copilot.