Gemini CLI is a capable agent in a window made of text. Add Filamental as an MCP server and what it reads and builds becomes a structure on your own disk, which you can look at in 3D or flat on the screen beside it, and change by hand when it gets something wrong. One entry in settings.json, free on every plan.
Add an mcpServers entry named filamental, running node against the bundled server with your space folder passed after a --vault argument:
{
"mcpServers": {
"filamental": {
"command": "node",
"args": [
"--no-warnings",
"/absolute/path/to/filamental-mcp/dist/index.js",
"--vault",
"/absolute/path/to/your/vault"
]
}
}
}Put it in settings.json in the .gemini folder of your home directory to have it everywhere, or in .gemini/settings.json inside a project to scope it to that project. Copy both absolute paths from Settings, then AI, then Other AI apps inside Filamental rather than typing them, since the server ships with the application and lands somewhere platform-specific.
Start Gemini CLI and run /mcp, or gemini mcp list from the shell, and Filamental should be listed as connected, with its tools.
A word on trust. Gemini CLI can skip the confirmation before each tool call for a server you mark as trusted. We’d leave that off until you’ve watched it work for a while, since the server can edit your space and changes made over MCP can’t be undone from inside Filamental.
A terminal is a good place to give instructions and a poor place to see a shape. When an agent builds something with forty parts and the relationships between them, the transcript is the least useful view of what it made.
With Filamental open beside it, the nodes appear as Gemini creates them, with their classifications and the typed connections between them, so you can see straight away when it has filed something in the wrong place or joined two things that shouldn’t be. Putting it right takes a drag, and the next thing Gemini reads is your correction.
Because the structure is typed, Gemini can follow it rather than search it. Ask how two things are connected and it reads back the route, the same one Pathfinder lights up on your screen.
A short briefing arrives when the connection opens, covering what nodes, categories and connectors are and the conventions that keep a structure readable, and the full skill guide and file format specification are pulled on demand only when a task needs them.
Both are published on this website in full, word for word as they are sent: the skill file and the file format reference. Nothing else runs underneath them. The server is a local process with no network access of its own, and Filamental has no account, no sign-in and no telemetry of any kind.
One prerequisite, and it is the only one that catches people out. The server reads the index Filamental builds and never builds one itself, so the space has to have been opened in the application at least once on that machine. A folder it has not seen stops with a message saying exactly that. If reading a folder somebody sent you is what you want, point the AI at the files directly instead, since a space is plain Markdown and no index is involved.
You need Node.js 22 or later and Filamental v0.2.4 or later, on Windows, macOS or Linux. The server ships inside the application, so there is nothing separate to install, and it is also on npm as filamental-mcp for a machine without Filamental on it.
The connection is free on the Personal plan, which is free permanently with no account and no card, and it is not metered, credited or seat-limited. The paid tier is $120 a year, or $12 a month, and covers handing work over beyond the one live link that’s free.
Add an mcpServers entry named filamental to settings.json, with command set to node and args holding the bundled server's path, then --vault and your space folder. Put it in settings.json in the .gemini folder of your home directory to have it everywhere, or in .gemini/settings.json inside a project to scope it there. Copy both absolute paths from Settings, then AI, then Other AI apps inside Filamental.
Run /mcp inside a Gemini CLI session, or gemini mcp list from the shell. Both show each configured server, whether it connected and the tools it offers. If Filamental is listed but will not connect, the usual cause is a space that has never been opened in the application on that machine, since the server reads the index Filamental builds and never builds one itself.
We would leave it off to begin with. Gemini CLI's trust setting skips the confirmation before every tool call, and the Filamental server can create, edit and connect nodes in your space, and changes made over MCP cannot be undone from inside Filamental. Once you have watched it work and are happy with what it does, turning it on is a reasonable convenience, and everything it changes is still in plain files you can read.
No. The MCP server is a separate local process reading the same folder, so it works whether or not the application window is open, and the --vault argument means it does not depend on which space you last had open. Many people keep Filamental open on a second screen anyway, since watching the structure change while the model writes to it is the quickest way to catch it misunderstanding something.
No, because there is nothing of ours in the path. The server is a local process talking to Markdown files in a folder on your own disk, and whatever leaves the machine is what Gemini CLI sends to Google, exactly as it does for the rest of your work. Filamental has no account, no sign-in and no telemetry of any kind.
Nothing. It is free on the Personal plan, which is free permanently with no account and no card, and it is not metered, credited or seat-limited. The paid tier, at $120 a year or $12 a month, covers handing work over beyond the one live link that's free, such as more links, updating a link already sent, the HTML download and presenting.
Then open the space beside the terminal and watch it arrive. Rearranging what it got wrong is a quick way to find out what you actually think.