Gemini CLI

Gemini CLI talks to the hosted server over streamable HTTP and signs in with OAuth, which it discovers from the server on its own. The extension is the one-step install: its gemini-extension.json adds the server, and its GEMINI.md is the same skill body every plugin carries — the standing instructions to fetch your profile before replying, stay inside its limits, and log the words used. Gemini CLI loads that file into context, so you get the instructions without a skill loader.

Extension

1. Install the extension from the langweave plugins repository, and confirm when Gemini CLI asks:

   gemini extensions install https://github.com/Jesusdiazrivero/langweave-plugins

2. Restart Gemini CLI — an installed extension takes effect in a new session.

3. Run /mcp auth langweave. Your browser opens lexmem’s sign-in page (lexmem keeps your langweave account); sign in and allow access.

To update it later, run gemini extensions update langweave.

Command

1. Add the server on its own, without the instructions file:

   gemini mcp add --transport http langweave https://api.langweave.tech/mcp

   That saves it for the current project (.gemini/settings.json). Add --scope user to have it in every project (~/.gemini/settings.json).

2. Start Gemini CLI and run /mcp auth langweave. Sign in in the browser and allow access.

Note

Use one or the other. If both are present, Gemini CLI merges them and your own settings win over the extension’s; gemini mcp remove langweave removes an entry you added with the command.

Check that it works

/mcp lists langweave as connected with five tools. Ask for something ordinary: it should call get_language_profile first and log_exposures after a reply that wove something. If the tools say you have not chosen a language, choose one in the app first.

The server also offers a language-mode prompt, which Gemini CLI lists as the slash command /language-mode. It turns weaving on explicitly for the rest of a conversation, with your profile attached.