Filamental and VS Code

The config that works everywhere else does nothing at all in VS Code

It reads the same local server as Claude and Cursor, over the same transport, but through a different key, and a file with the wrong one is ignored rather than rejected. Here’s the shape it actually wants, and what having a structure beside the code is worth to the model in your editor. Free on every plan.

Connecting it

A servers key, and an explicit stdio type

Put this in .mcp.json or .vscode/mcp.json at the root of your project:

{
  "servers": {
    "filamental": {
      "type": "stdio",
      "command": "node",
      "args": [
        "--no-warnings",
        "/absolute/path/to/filamental-mcp/dist/index.js",
        "--vault",
        "/absolute/path/to/your/vault"
      ]
    }
  }
}

Two differences from the block every other client takes, and both matter. The top-level key is servers rather than mcpServers, and each entry declares "type": "stdio". Copy both absolute paths from Settings, then AI, then Other AI apps inside Filamental, which prints the VS Code shape separately for this reason.

Get the key wrong and nothing is reported. The file parses, the entry is ignored, and the server simply never appears in the tool list, which is a frustrating half hour if you don’t know to look for it. If a connection you’re certain about hasn’t turned up, check that word first.

The version that survives a commit

Absolute paths don’t travel, and a relative one does

The block above carries two paths that are specific to your machine, so committing it hands everyone else a file that can’t work. This one travels:

{
  "servers": {
    "filamental": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "filamental-mcp", "--vault", "./knowledge"]
    }
  }
}

npx fetches the server from npm rather than pointing at your own install, and a relative vault path is resolved against the folder the editor opened, so ./knowledge means the same thing on every machine that clones the repository.

It doesn’t remove the prerequisite below, only the path editing. It also wants Node on the path and is a shade slower to start, so for your own machine the direct version is still the better one.

The one prerequisite

The space has to have been opened in Filamental once

The server reads the index Filamental builds and never builds one itself, so a folder the application hasn’t opened on that machine can’t be reached through MCP. It stops with a message saying exactly that rather than failing obscurely, which is at least the right way round.

That makes MCP a route for people who have the application, rather than a way to browse somebody else’s folder without it. If reading a folder you were sent is what you actually want, point the AI at the files directly, since a space is plain Markdown and the structure is in the front matter, so no index is involved at all.

What it adds

Relationships that are stated rather than inferred

An editor with a model in it is already good at what calls what, since that’s in the code and can be found. What isn’t in the code is why a boundary sits where it does, which constraint ruled out the obvious approach, or which decision a proposed change quietly reverses. In a space those are typed connections between named things, so the model follows them instead of guessing.

It works for Copilot as readily as for anything else. A server configured here is available to agent mode, and nothing about ours is tied to a particular assistant: it speaks stdio and doesn’t know which model is on the other end.

Writing works as well as reading, so a decision can be recorded from the same window while it’s being made, rather than reconstructed later by whoever inherits it.

What we tell your AI

You can read the exact text it’s handed

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.

Requirements, and what it costs

Node 22, and nothing

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’s nothing separate to install, and it is also on npm as filamental-mcp.

The connection is free on the Personal plan, which is free permanently with no account and no card, and it isn’t metered, credited or seat-limited. The paid tier is $120 a year, or $12 a month and buys sending a finished structure to somebody who doesn’t have the application.

Asked before connecting

Six straight answers

How do I add Filamental as an MCP server in VS Code?

Put an entry named filamental in .mcp.json or .vscode/mcp.json at the root of your project, under a top-level servers key, with the type set to stdio, the command set to node, and your space folder passed after a --vault argument. Copy both absolute paths for your machine from Settings, then AI, then Other AI apps inside Filamental, which prints the VS Code shape separately from the others.

Why does the config that works in Cursor do nothing in VS Code?

Because VS Code uses a different top-level key. Claude Desktop, Claude Code, Cursor and LM Studio all read mcpServers, while VS Code reads servers and additionally wants an explicit type of stdio on each entry. A file carrying the wrong key is ignored rather than rejected, with no error anywhere, so the usual symptom is a server that simply never appears in the tool list.

Does this work with GitHub Copilot specifically?

Yes. MCP servers configured in VS Code are available to Copilot's agent mode, so the same connection that gives the editor access to your space gives Copilot access to it. Nothing about the server is Claude-specific: it speaks stdio, which is the standard local transport, and it neither knows nor cares which model is on the other end of the conversation.

Do I need Filamental installed to use the MCP server?

Yes, and the space has to have been opened in the application at least once on that machine. The server reads the index Filamental builds and never builds one itself, so pointing it at a folder that hasn’t been opened stops with a message saying exactly that. If you only want an AI to read a folder somebody sent you, point it at the files directly instead, since a space is plain Markdown and needs no index at all.

Can I commit the VS Code MCP config to the repository?

Only if you use the portable form, because the ordinary one carries two absolute paths that are specific to your own machine. Setting command to npx with the arguments -y and filamental-mcp fetches the server from npm rather than from your install, and a relative vault path such as ./knowledge is resolved against the folder the editor opened, so both halves travel. Everyone cloning it still has to open that folder in Filamental once.

What does the VS Code connection cost?

Nothing. It is free on the Personal plan, which is free permanently with no account and no card, and it isn’t metered, credited or seat-limited, because there’s no server of ours in the path to meter it at. Publishing one link is free too, and the paid tier, at $120 a year or $12 a month, adds more links, updates to a link already sent, the HTML download and presenting.

A first thing to try

Ask it why the part of the codebase you like least is like that

Then keep the answer as a structure rather than as a message you’ll lose. It takes about as long as writing the commit, and it’s the version that survives you.

Platforms
Win / macOS / Linux
Personal
Free, permanent
Professional
$120 / year, or $12 / month