Letronna®
GitHub ↗

Verbinden und erweitern

MCP servers

Giving Letronna more tools from MCP servers: adding them, signing in, and choosing which tools it may use.

Diese Seite ist noch nicht übersetzt und wird deshalb auf Englisch gezeigt. Du kannst sie auf GitHub übersetzen. GitHub ↗

An MCP server gives Letronna tools of its own: a database, an issue tracker, your notes. Letronna reaches servers that run as a command on your computer, and servers reached by URL. They are listed in mcp.json in your data folder, in the same shape Claude Desktop and Cursor use, so you can copy a list across.

Adding a server

From the terminal:

letronna mcp add files -- npx -y @modelcontextprotocol/server-filesystem ~/notes
letronna mcp add github -e GITHUB_TOKEN=... -- npx -y @modelcontextprotocol/server-github
letronna mcp add --transport http sentry https://mcp.sentry.dev/mcp
Option for addDoes
-t, --transport <kind>stdio (a command run here) or http; a URL means http
-e, --env <KEY=value>an environment variable for the command
-H, --header <"K: v">a header sent to a server reached by URL
--client-id <id>the app id a server that registers no apps gave you
--callback-port <port>the port that app id was registered with
CommandDoes
letronna mcp listevery server, started to see it answer
letronna mcp get <name>one server in full
letronna mcp remove <name>take one out
letronna mcp login <name>sign in to a server reached by URL, in the browser
letronna mcp logout <name>forget that sign-in

In the window, the MCP tab on the Capabilities page lists the servers, with a switch for each tool.

Writing mcp.json yourself

{
  "mcpServers": {
    "everything": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-everything"],
      "autoApprove": ["echo"]
    },
    "hosted": {
      "url": "https://example.com/mcp",
      "headers": { "Authorization": "Bearer ${HOSTED_KEY}" },
      "excludeTools": ["delete_*"]
    }
  }
}
  • Every call asks first, unless the tool is in autoApprove. "autoApprove": "readOnly" trusts the server's own word on which tools only read.
  • includeTools and excludeTools choose which tools Letronna sees.
  • Any text can use ${VAR}, or ${VAR:-default} for a value when the variable is not set.
  • "disabled": true leaves a server off.
  • cwd sets the folder a command runs in, and timeout_ms and startup_timeout_ms how long Letronna waits.

Servers reached by URL use streamable HTTP; the older SSE transport is not supported. Servers restart when they stop, and the list is read again when mcp.json changes.

In a session

/mcp
/mcp reload
/mcp reconnect github

/mcp shows the servers and how each is doing. It also takes enable, disable, login, logout and log (what a server printed). A server's prompts run as slash commands, and its resources can be read.

When servers bring many tools, Letronna looks them up when needed instead of sending all of them with every request.

Letronna as an MCP server

Another agent can use Letronna's media and search tools over MCP:

letronna mcp serve

--yes runs tools that need a yes without one, and --allow-turn also offers a whole Letronna turn as a tool.

Diese Seite auf GitHub bearbeiten ↗
ZurückTelegram, Discord and SlackWeiterPlugins, hooks and more