Command Line
CLI MCP Server
Run the Model Context Protocol server so an AI client can read and edit your snippets.
Snipset ships a Model Context Protocol (MCP) server so an AI client can read and edit your snippets, notes, tasks, and other data through the CLI.
What MCP is
MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems such as local files, databases, and tools. An AI client (the host) speaks MCP to a server that exposes data and actions. Snipset is one such server.
Run the server
snipset mcp
This runs the MCP server over stdio: it reads requests on standard input and writes responses on standard output. You do not run it interactively yourself; an MCP client launches it as a subprocess and communicates over those streams. Point it at a specific database with --db <PATH> if you do not want the default.
Connect an MCP client
An MCP client such as opencode can launch snipset mcp as a local (stdio) server. The Linux installer can write an opencode MCP configuration for you with the --with-mcp flag:
bash install-snipset.sh --version v0.2.0 --with-mcp
This writes ~/.config/opencode/opencode.json only when the file does not already exist; it never overwrites an existing configuration. If you already have an opencode configuration, add a Snipset server entry to it by hand that runs the snipset mcp command.
Which operations work
The MCP server uses the same database access as the rest of the CLI, so database-only operations work whether or not Snipset Desktop is running, while bridge-dependent operations (see the command reference) need the desktop app. Keep the app open if you want the client to use timer, generation, or parsing features.