FxFiles × Model Context Protocol

Take Your AI's Memory Anywhere

Give your AI assistant a private, encrypted second brain that lives on the decentralized Fula network — readable and writable from any tool, on any device. Your knowledge follows you, not the vendor.

Get FxFiles Connect your AI

Your AI forgets everything between sessions

Every new chat starts from zero. It doesn't know your project, your past decisions, or the notes you took last week — so it guesses, and generic guesses waste your time. A second brain fixes that: a structured knowledge base your AI reads before it answers and writes back to as you work. FxFiles is where that brain can safely live.

How it works

Three steps. About five minutes. No servers to run.

1

Create a private group

In FxFiles, open Shared → Collaborate and create an encrypted collaboration group — call it brain. This is the only folder your AI will ever touch.

2

Connect your AI

Point Claude or ChatGPT at the Fula MCP server — a hosted one-liner, or a local server if you want every byte to stay on your machine — then authorize the group.

3

It reads & writes

Your AI now opens every session already knowing what you know, and files new decisions back into the group as you work. Encrypted, portable, yours.

A memory you actually own

Not a vendor's chat-history feature. A knowledge base under your control.

Encrypted end-to-end

Everything is encrypted on your device before it leaves. The network stores ciphertext it can never read.

You hold the keys

Your master key never leaves your device. No cloud vendor can read, mine, or lock you out of your own knowledge.

Portable across devices

The same brain is reachable from your laptop, your phone, and every AI tool you use — it isn't trapped on one machine.

Works with your AI

It speaks the Model Context Protocol that Claude and ChatGPT already understand. No plugin to build, no connector to babysit.

Scoped & revocable

The AI sees only the one group you shared and can't permanently delete your files. Revoke its access whenever you want.

Decentralized

Your brain is backed by the Fula network, not a single provider — so it survives outages, account bans, and vendor shutdowns.

A small, boring set of operations

Your AI just gets file operations it already understands — that's the whole surface.

fula_store_filewrite a file
fula_read_fileread one, by path or id
fula_list_fileslist what's there
fula_searchfind by name or path
fula_create_folderorganize into folders
fula_remove_fileremove an entry
fula_tag_filetag a file
fula_list_tagslist what's tagged

Connect it in under five minutes

Two ways in. Take the easy one unless you have a reason not to.

Easiest · nothing to install

Hosted connector

Point your AI at the hosted Fula MCP and sign in with the same Google account you use for FxFiles. In Claude Code, it's one line:

claude mcp add --transport http fula \
  https://mcp.cloud.fx.land/mcp

Then run /mcp to authorize. In Claude Desktop, Claude web, or ChatGPT, add a custom connector for https://mcp.cloud.fx.land/mcp.

Private · everything on-device

Local server

Run the MCP on your own machine so every byte stays local. It needs Node and fetches a small verified binary on first run — no npm install:

npx -y @functionland/fula-mcp

Then copy the config from FxFiles → Share with AI Agent → Desktop into your client (it carries a scoped capability — treat it like an API key).

Finally, pair it to your group: ask your AI "what is your Fula identity?", then in FxFiles open your group → Share with AI Agent → paste the FULA-… id → Authorize. Done.

You stay in control

  • Your master key never leaves your device — the encryption enforces it, not a policy.
  • The AI can see and add to only the one group you shared, nothing else in your account.
  • It cannot permanently delete your files.
  • You can revoke access at any time.

Give your AI a memory that's actually yours

Install FxFiles, create a group, and connect your AI. Your next session will already know your work.

Get FxFiles Open FxFiles Web